Detecting Credit Card Fraud using Periodic Features

Size: px
Start display at page:

Download "Detecting Credit Card Fraud using Periodic Features"

Transcription

1 Detectng Credt Card Fraud usng Perodc Features Alejandro Correa Bahnsen, Djamla Aouada, Aleksandar Stojanovc and Björn Ottersten Interdscplnary Centre for Securty, Relablty and Trust Unversty of Luxembourg, Luxembourg Emal: Abstract When constructng a credt card fraud detecton model, t s very mportant to extract the rght features from transactonal data. Ths s usually done by aggregatng the transactons n order to observe the spendng behavoral patterns of the customers. In ths paper we propose to create a new set of features based on analyzng the perodc behavor of the tme of a transacton usng the von Mses dstrbuton. Usng a real credt card fraud dataset provded by a large European card processng company, we compare state-of-the-art credt card fraud detecton models, and evaluate how the dfferent sets of features have an mpact on the results. By ncludng the proposed perodc features nto the methods, the results show an average ncrease n savngs of 3%. The aforementoned card processng company s currently ncorporatng the methodology proposed n ths paper nto ther fraud detecton system. Keywords Fraud detecton; von Mses dstrbuton; Costsenstve learnng I. INTRODUCTION Credt card fraud has been a growng problem worldwde. Durng 202 the total level of fraud reached.33 bllon Euros n the Sngle Euro Payments Area, whch represents an ncrease of 4.8% compared wth 20 []. Moreover, payments across non tradtonal channels (moble, nternet,...) accounted for 60% of the fraud, whereas t was 46% n Ths opens new challenges as new fraud patterns emerge, and current fraud detecton systems are less successful n preventng these frauds. Furthermore, fraudsters constantly change ther strateges to avod beng detected, somethng that makes tradtonal fraud detecton tools, such as expert rules, nadequate [2]. The use of machne learnng n fraud detecton has been an nterestng topc n recent years. Dfferent detecton systems that are based on machne learnng technques have been successfully used for ths problem, n partcular: neural networks [3], Bayesan learnng [4], artfcal mmune systems [5], hybrd models [6], support vector machnes [7], peer group analyss [8], onlne learnng [9] and socal network analyss [2]. When constructng a credt card fraud detecton model, t s very mportant to use those features that allow accurate classfcaton. Typcal models only use raw transactonal features, such as tme, amount, place of the transacton. However, these approaches do not take nto account the spendng behavor of the customer, whch s expected to help dscover fraud patterns [5]. A standard way to nclude these behavoral spendng patterns s proposed n [0], where Whtrow et al. proposed a transacton aggregaton strategy n order to take nto account a customer spendng behavor. The computaton of the aggregated features conssts n groupng the transactons made durng the last gven number of hours, frst by card or account number, then by transacton type, merchant group, country or other, followed by calculatng the number of transactons or the total amount spent on those transactons. In ths paper, we propose a new set of features based on analyzng the tme of a transacton. The logc behnd t s that a customer s expected to make transactons at smlar hours. We, hence, propose a new method for creatng features based on the perodc behavor of a transacton tme, usng the von Mses dstrbuton []. In partcular, these new tme features should estmate f the tme of a new transacton s wthn the confdence nterval of the prevous transacton tme. Furthermore, usng a real credt card fraud dataset provded by a large European card processng company, we compare the dfferent sets of features (raw, aggregated and perodc), usng two knds of classfcaton algorthms; cost-nsenstve [2] and example-dependent cost-senstve [3]. The results show an average ncrease n the savngs of 3% by usng the proposed perodc features. Addtonally, the outcome of ths paper s beng currently used to mplement a state-of-the-art fraud detecton system, that wll help to combat fraud once the mplementaton stage s fnshed. The remander of the paper s organzed as follows. In Secton 2, we dscuss current approaches to create the features used n fraud detecton models. Then, n Secton 3, we present our proposed methodology to create perodc features. Afterwards, the expermental setup and the results are gven n Sectons 4 and 5. Fnally, conclusons and dscussons of the paper are presented n Secton 6. II. TRANSACTION AGGREGATION STRATEGIES When constructng a credt card fraud detecton algorthm, the ntal set of features (raw features) nclude nformaton regardng ndvdual transactons. It s observed throughout the lterature, that regardless of the study, the set of raw features s qute smlar. Ths s because the data collected durng a credt card transacton must comply wth nternatonal fnancal reportng standards. In TABLE I, the typcal credt card fraud detecton raw features are summarzed. Several studes use only the raw features n carryng ther analyss [3], [4]. However, as noted n [4], a sngle transacton nformaton s not suffcent to detect a fraudulent transacton, snce usng only the raw features leaves behnd mportant nformaton such as the consumer spendng behavor, whch s usually used by commercal fraud detecton systems [0]. To deal wth ths, n [5], a new set of features were proposed such that the nformaton of the last transacton made wth the same credt card s also used to make a predcton.

2 TABLE I. SUMMARY OF TYPICAL RAW CREDIT CARD FRAUD DETECTION FEATURES Attrbute name Descrpton Transacton ID Transacton dentfcaton number Tme Date and tme of the transacton Account number Identfcaton number of the customer Card number Identfcaton of the credt card Transacton type e. Internet, ATM, POS,... Entry mode e. Chp and pn, magnetc strpe,... Amount Amount of the transacton n Euros Merchant code Identfcaton of the merchant type Merchant group Merchant group dentfcaton Country Country of trx Country 2 Country of resdence Type of card e. Vsa debt, Mastercard, Amercan Express... Gender Gender of the card holder Age Card holder age Bank Issuer bank of the card The objectve, s to be able to detect very dssmlar contnuous transactons wthn the purchases of a customer. The new set of features nclude: tme snce the last transacton, prevous amount of the transacton, prevous country of the transacton. Nevertheless, these features do not take nto account consumer behavor other than the last transacton made by a clent, ths leads to havng an ncomplete profle of customers. A more compressve way to take nto account a customer spendng behavor s to derve some features usng a transacton aggregaton strategy. Ths methodology was ntally proposed n [0]. The dervaton of the aggregaton features conssts n groupng the transactons made durng the last gven number of hours, frst by card or account number, then by transacton type, merchant group, country or other, followed by calculatng the number of transactons or the total amount spent on those transactons. Ths methodology has been used by a number of studes [7] [9], [5] [8]. When aggregatng a customer transactons, there s an mportant queston on how much to accumulate, n the sense that the margnal value of new nformaton may dmnsh as tme passes. [0] dscuss that aggregatng 0 transactons s not lkely to be more nformatve than aggregatng 00 transactons. Indeed, when tme passes, nformaton lose ther value, n the sense that a customer spendng patterns are not expected to reman constant over the years. In partcular, Whtrow et al. defne a fxed tme frame to be 24, 60 or 68 hours. Let S be a set of N transactons,.e., N = S, where each transacton s represented by the feature vector x = [x, x2,..., xk ], where k s the number of features, and labelled usng the class label y {0, }. Then, the process of aggregatng features conssts n selectng those transactons that were made n the prevous t p hours, for each transacton n the dataset S, { S agg T RX agg (S,, t p ) = x amt ( l x d l = x d ) ( hours(x tme, x tme l ) < t p ) } N l=, () where T RX agg s a functon that creates a subset of S assocated wth a transacton wth respect to the tme frame t p, N = S, beng the cardnalty of a set, x tme s the tme of transacton, x amt s the amount of transacton, the customer dentfcaton number of transacton, and x d TABLE II. EXAMPLE CALCULATION OF AGGREGATED FEATURES. WHERE, x a IS THE NUMBER OF TRANSACTIONS IN THE LAST 24 HOURS AND x a2 IS THE SUM OF THE TRANSACTIONS AMOUNTS IN THE SAME TIME PERIOD. Raw features Agg. features TrxId CardId Tme Type Country Amt. x a x a2 0/0 8:20 POS Lux /0 20:35 POS Lux /0 22:30 ATM Lux /0 00:50 POS Ger /0 9:8 POS Ger /0 23:45 POS Ger /0 06:00 POS Lux hours(t, t 2 ) s a functon that calculates the number of hours between the tmes t and t 2. Afterwards the feature number of transactons and amount of transactons n the last t p hours are calculated as: x a = S agg, (2) and respectvely. x a2 = x amt S agg x amt, (3) To further clarfy how the aggregated features are calculated we show an example. Consder a set of transactons made by a clent between the frst and thrd of January of 205, as shown n TABLE II. Then we estmate the aggregated features (x a and x a2 ) by settng t p = 24 hours. Moreover, the total number of aggregated features can grow qute quckly, as t p can have several values, and the combnaton of combnaton crtera can be qute large as well. In [7], we used a total of 280 aggregated features. In partcular we set the dfferent values of t p to:, 3, 6, 2, 8, 24, 72 and 68 hours. Then calculate the aggregated features usng () wth the followng groupng crtera: country, type of transacton, entry mode, merchant code and merchant group. III. PROPOSED PERIODIC FEATURES When usng the aggregated features, there s stll some nformaton that s not completely captured by those features. In partcular we are nterested n analyzng the tme of the transacton. The logc behnd ths, s that a customer s expected to make transactons at smlar hours. The ssue when dealng wth the tme of the transacton, specfcally, when analyzng a feature such as the mean of transactons tme, s that t s easy to make the mstake of usng the arthmetc mean. Indeed, the arthmetc mean s not a correct way to average tme because, as shown n Fg., t does not take nto account the perodc behavor of the tme feature. For example, the arthmetc mean of transacton tme of four transactons made at 2:00, 3:00, 22:00 and 23:00 s 2:30, whch s counter ntutve snce no transacton was made close to that tme. We propose to overcome ths lmtaton by modelng the tme of the transacton as a perodc varable, n partcular usng the von Mses dstrbuton []. The von Mses dstrbuton, also known as the perodc normal dstrbuton, s a dstrbuton of a wrapped normal dstrbuted varable across a crcle. The von Mses probablty dstrbuton of a set of examples

3 Fg.. Analyss of the tme of a transacton usng a 24 hour clock. The arthmetc mean of the transactons tme (dashed lne) do not accurately represents the actual tmes dstrbuton. Fg. 2. Ftted von Mses dstrbuton ncludng the perodc mean (dashed lne) and the probablty dstrbuton (purple area). D = {t, t2,, tn } for a gven angle tx s gven by f (tx µvm, σvm ) = e σvm cos(tx µvm ) 2πI0 σvm (4) where I0 (κ) s the modfed Bessel functon of order 0, and µvm and σvm are the perodc mean and perodc standard devaton, respectvely. In Appendx A we present the calculaton of µvm and σvm. In partcular, we are nterested n calculatng a confdence nterval (CI) for the tme of a transacton. For dong that, ntally we select a set of transactons made by the same clent n the last tp hours (Sper ) as: tme d Sper T RXvM (S,, tp ) = xl xd l = x N tme hours(xtme, x ) < t. (5) p l l= Afterwards, the probablty dstrbuton functon of the tme of the set of transactons s calculated as:. (6) xtme vonmses µ (S ), vm per σvm (Sper ) In Fg. 2, the von Mses dstrbuton calculaton for the earler example s shown. It s observed that the arthmetc mean s dfferent from the perodc mean, the latter beng a more realstc representaton of the actual transactonal tmes. Then, usng the estmated dstrbuton, a new set of features can be extracted,.e., a bnary feature (xp ) f a new transacton tme s wthn the confdence nterval range wth probablty α. An example s presented n Fg. 3. Furthermore, other features can be calculated, as the confdence nterval range can be calculated for several values of α, and also the tme perod can have an arbtrary sze. Fg. 3. Expected tme of a transacton (green area). Usng the confdence nterval, a transacton can be flag normal or suspcous, dependng whether or not the tme of the transacton s wthn the confdence nterval. Addtonally, followng the same example presented n TABLE II, we calculate a feature xp, as a bnary feature that takes the value of one f the current tme of the transacton s wthn the confdence nterval of the tme of the prevous transactons wth a confdence of α = 0.9. The example s shown n TABLE III, where the arthmetc and perodc means dffer, as for the last transacton. Moreover, the new feature helps to get a better understandng of when a customer s expected to make transactons. Fnally, when calculatng the perodc features, t s mportant to use longer tme frames tp, snce f the dstrbuton s calculated usng only a couple of transactons t may not be as relevant of a customer behavor patterns, compared aganst usng a full year of transactons. Evdently, f tc s less than 24 hours, any transacton made afterwards wll not be expected

4 TABLE III. EXAMPLE CALCULATION OF PERIODIC FEATURES. WHERE x p IS A BINARY FEATURE THAT INFORMS WHENEVER A TRANSACTION IS BEING MADE WITHIN THE CONFIDENCE INTERVAL OF THE TIME OF THE TRANSACTIONS. Raw features Arthmetc Perodc features Id Tme mean mean Confdence nterval x p 0/0/5 8:20 2 0/0/5 20:35 3 0/0/5 22:30 9:27 9:27 5:45-23:0 True 4 02/0/5 00:50 20:28 20:28 7:54-23:03 False 5 02/0/5 9:8 6:34 22:34 8:5-00:7 True 6 02/0/5 23:45 6:9 2:07 5:2-02:52 True 7 03/0/5 06:00 8:33 22:33 7:9-0:46 False to be wthn the dstrbuton of prevous transactonal tmes. To avod ths, we recommend usng at least the prevous 7 days of transactonal nformaton, therefore, havng a better understandng of ts behavoral patterns. IV. EXPERIMENTAL SETUP In ths secton, frst the dataset used for the experments s descrbed. Second, the evaluaton measure used to compare the algorthms s shown. Lastly, the algorthms used n ths paper are brefly explaned. A. Database For ths paper we used a dataset provded by a large European card processng company. The dataset conssts of fraudulent and legtmate transactons made wth credt and debt cards between January 202 and June 203. The total dataset contans 20,000,000 ndvdual transactons, each one wth 27 attrbutes, ncludng a fraud label ndcatng whenever a transacton s dentfed as fraud. Ths label was created nternally n the card processng company, and can be regarded as hghly accurate. In the dataset only 40,000 transactons were labeled as fraud, leadng to a fraud rato of 0.025%. Furthermore, usng the methodologes for feature extracton descrbed n Secton II, we estmate a total of 293 features. Also, for the experments, a smaller subset of transactons wth a hgher fraud rato, correspondng to transactons made wth magnetc strpe, s selected. Ths dataset contans 236,735 transactons and a fraud rato of.50%. In ths dataset, the total fnancal losses due to fraud are 895,54 Euros. Ths dataset was selected because t s the one where most frauds occur. The total dataset s dvded nto 3 subsets: tranng, valdaton and testng. Each one contanng 50%, 25% and 25% of the transactons respectvely. TABLE IV summarzes the dfferent datasets. B. Evaluaton measure When evaluatng a credt card fraud detecton model, typcally a standard bnary classfcaton measure, such as msclassfcaton error, recever operatng characterstc (ROC), Kolmogorov-Smrnov (KS) or F Score statstcs, s used [9], [9], [20]. However, these measures may not be the most approprate evaluaton crtera when evaluatng fraud detecton models, because they tactly assume that msclassfcaton errors carry the same cost, smlarly wth the correct classfed transactons. Ths assumpton does not hold n practce, when wrongly predctng a fraudulent transacton as legtmate TABLE IV. SUMMARY OF THE DATASETS Set Transactons %Frauds Cost Total 236, ,54 Tranng 94, ,078 Valdaton 70, ,90 Testng 7, ,67 TABLE V. CREDIT CARD FRAUD COST MATRIX [7] Predcted Postve c = Predcted Negatve c = 0 Actual Postve Actual Negatve y = y = 0 C T P = C a C F P = C a C F N = Amt C T N = 0 carres a sgnfcantly dfferent fnancal cost than the nverse case. In order to take nto account the dfferent costs of fraud detecton durng the evaluaton of an algorthm, n [7], we proposed a cost matrx [3] that takes nto account the actual example-dependent fnancal costs. In TABLE V, the cost matrx s presented, where the predcton of the algorthm c s a functon of the k features of transacton, x = [x, x2,..., xk ], y s the true class of the transacton, and the cost assocated wth two types of correct classfcaton, namely, true postves C T P, and true negatves C T N ; and the two types of msclassfcaton errors, namely, false postves C F P, and false negatves C F N, are presented. Moreover, our cost matrx defnes the cost of a false negatve to be the amount of the transacton Amt, and the costs of false postve and true postve to be the admnstratve cost C a related to analyzng the transacton and contactng the card holder. Afterwards, usng the example-dependent cost matrx, a cost measure s calculated takng nto account the actual costs of each transacton. Let S be a set of N transactons,.e., N = S, where each transacton s represented by the augmented feature vector x = [x, C T P, C F P, C F N, C T N ], and labelled usng the class label y {0, }. A classfer f whch generates the predcted label c for each transacton, s traned usng the set S. Then the cost of usng f on S s calculated by Cost(f(S)) = N y ( c )Amt + c C a. (7) = Lastly, n order to have a measure that s easy to nterpret, we used the fnancal savngs as we defned n [2]. In partcular, the savngs measure we proposed a savngs measure that compare the cost of an algorthm versus the cost of usng no algorthm at all. In the case of credt card fraud the cost of usng no algorthm s equal to the sum of the amounts of the fraudulent transactons N = y Amt. Then, the savngs are calculated as: Savngs(f(S)) = N = y c Amt c C a N = y Amt. (8) In other words, the sum of the amounts of the corrected predcted fraudulent transactons mnus the admnstratve cost ncurred n detectng them, dvded by the sum of the amounts of the fraudulent transactons.

5 Fg. 4. Comparson of the dfferent algorthms, traned wth only the raw features (raw), only the aggregated features (agg) and both (raw + agg). In average, by usng both the raw and the aggregated features the savngs are doubled. Fg. 5. Comparson of the proposed perodc (per) set of features. It s observed, that when the new set of features are combned wth the aggregated features, an addtonal ncrease of savngs of 6.4% s made. C. Algorthms For the experments we used three cost-nsenstve classfcaton algorthms: decson tree (DT ), logstc regresson (LR) and a random forest (RF ), usng the mplementaton of Scktlearn [22]. Furthermore, n prevous works, we have shown the need to use algorthms that take nto account the dfferent costs assocated wth fraud detecton [8]. In partcular we also used three cost-senstve algorthms, namely, Bayes mnmum rsk (BM R) [7], [8], cost-senstve logstc regresson (CSLR) [23] and cost-senstve decson tree (CSDT ) [2]. The BM R s a decson model based on quantfyng tradeoffs between varous decsons usng probabltes and the costs that accompany such decsons. In the case of credt card fraud detecton, a transacton s classfed as fraud f C a Amt ˆp, and as legtmate f false. Where ˆp s the estmated probablty of a transacton beng fraud gven x. The other two cost-senstve methods, CSLR and CSDT, are based on ntroducng the example-dependent costs nto a logstc regresson and a decson tree algorthm, by changng the objectve functon of the models to one that s cost-senstve. For a further dscusson see [23] and [2], respectvely. The mplementaton of the cost-senstve algorthms s done usng the CostCla lbrary. Moreover, each algorthm was traned, usng the dfferent sets of features: raw features (raw) as shown n TABLE I, aggregated features (agg) usng equatons (2) and (3), and the perodc features (per) as descrbed n Secton III. V. RESULTS Frst we evaluate the savngs of the dfferent algorthms usng only the raw features (raw), only the aggregated features (agg) and both (raw + agg). The results are shown n Fg. 4. Note that all the algorthms generate savngs,.e., no algorthm performs worse than usng no algorthm at all. The CSDT algorthm s the one that performs best, n partcular when usng both the raw and aggregated features. When analyzng the results usng the dfferent set of features, the aggregated features perform better than usng only the raw features n Fg. 6. Comparson of the average ncrease n savngs when ntroducng each set of features compared wth the results of usng only the raw set of features. all the cases. Ths confrms the ntuton of the need of usng the customer behavor patterns n order to dentfy fraudulent transactons. On average, by usng both the raw and the aggregated features the savngs are doubled. Then, we evaluate the results of the perodc set of features. In Fg. 5, the results are shown. The new set of perodc features ncrease the savngs by an addtonal 3%. The algorthm wth the hghest savngs s the CSDT, closely followed by CSLR. Smlarly to usng the extended aggregated features, the perodc features do not perform well when used only wth raw features. It s when combned the set of aggregated features that an ncrease n savngs s found. Fnally, n Fg. 6, we compare the average ncrease n savngs when ntroducng each set of features compared wth the results of usng only the set of raw features. Frst, the aggregated features gve an average ncrease n savngs of 20%. As prevously shown, n order to mprove the results, these new features need to be combned wth the aggregated features n order to ncrease savngs. Lastly, when combnng the prevous features wth the perodc features, the results ncrease by 287% compared wth usng raw features only.

6 VI. CONCLUSION AND DISCUSSION In ths paper we have shown the mportance of usng features that analyze the consumer behavor of ndvdual card holders when constructng a credt card fraud detecton model. We show that by preprocessng the data n order to nclude the recent consumer behavor, the performance ncreases by more than 200% compared to usng only the raw transacton nformaton. Moreover, we extended the current approaches to analyze the consumer behavor by proposng a new method to analyze the perodc behavor of the tme of a transacton usng the von Mses dstrbuton. The new proposed set of features ncreases the performance by 287%. However, because ths study was done usng a dataset from a fnancal nsttuton, we were not able to deeply dscuss the specfc features created, and the ndvdual mpact of each feature. Nevertheless, our framework s ample enough to be recreated wth any knd of transactonal data. Furthermore, when mplementng ths framework on a producton fraud detecton system, questons regardng response and calculaton tme of the dfferent features should be addressed. In partcular, snce there s no lmt on the number of features that can be calculated, a system may take too long to make a decson based on the tme spent recalculatng the features wth each new transacton. ACKNOWLEDGMENT Fundng for ths research was provded by the Fonds Natonal de la Recherche, Luxembourg, grant number AFR-PhD REFERENCES [] European Central Bank, Thrd report on card fraud, European Central Bank, Tech. Rep., 204. [2] V. Van Vlasselaer, C. Bravo, O. Caelen, T. Elass-Rad, L. Akoglu, M. Snoeck, and B. Baesens, APATE: A Novel Approach for Automated Credt Card Transacton Fraud Detecton usng Network- Based Extensons, Decson Support Systems, vol. 75, pp , 205. [3] R. Brause, T. Langsdorf, and M. Hepp, Neural data mnng for credt card fraud detecton, Proceedngs th Internatonal Conference on Tools wth Artfcal Intellgence, pp , 999. [4] S. Pangrah, A. Kundu, S. Sural, and A. Majumdar, Credt card fraud detecton: A fuson approach usng Dempster Shafer theory and Bayesan learnng, Informaton Fuson, vol. 0, no. 4, pp , Oct [5] S. Bachmayer, Artfcal Immune Systems, Artfcal Immune Systems, vol. 532, pp. 9 3, [6] M. Krvko, A hybrd model for plastc card fraud detecton systems, Expert Systems wth Applcatons, vol. 37, no. 8, pp , Aug [7] S. Bhattacharyya, S. Jha, K. Tharakunnel, and J. C. Westland, Data mnng for credt card fraud: A comparatve study, Decson Support Systems, vol. 50, no. 3, pp , Feb. 20. [8] D. J. Weston, D. J. Hand, N. M. Adams, C. Whtrow, and P. Juszczak, Plastc card fraud detecton usng peer group analyss, Advances n Data Analyss and Classfcaton, vol. 2, no., pp , Mar [9] A. D. Pozzolo, O. Caelen, Y.-A. Le Borgne, S. Waterschoot, and G. Bontemp, Learned lessons n credt card fraud detecton from a practtoner perspectve, Expert Systems wth Applcatons, vol. 4, no. 0, pp , Aug [0] C. Whtrow, D. J. Hand, P. Juszczak, D. J. Weston, and N. M. Adams, Transacton aggregaton as a strategy for credt card fraud detecton, Data Mnng and Knowledge Dscovery, vol. 8, no., pp , Jul [] N. I. Fsher, Statstcal Analyss of Crcular Data, 996, vol. 9. [2] C. M. Bshop, Pattern Recognton and Machne Learnng, ser. Informaton scence and statstcs. Sprnger, 2006, vol. 4, no. 4. [3] C. Elkan, The Foundatons of Cost-Senstve Learnng, n Seventeenth Internatonal Jont Conference on Artfcal Intellgence, 200, pp [4] R. Bolton and D. J. Hand, Unsupervsed proflng methods for fraud detecton, n Credt Scorng and Credt Control VII, 200. [5] D. Tasouls and N. Adams, Mnng nformaton from plastc card transacton streams, n Proceedngs n 8th Internatonal Conference on Computatonal Statstcs, [6] S. Jha, M. Gullen, and J. Chrstopher Westland, Employng transacton aggregaton strategy to detect credt card fraud, Expert Systems wth Applcatons, vol. 39, no. 6, pp , 202. [7] A. Correa Bahnsen, A. Stojanovc, D. Aouada, and B. Ottersten, Cost Senstve Credt Card Fraud Detecton Usng Bayes Mnmum Rsk, n 203 2th Internatonal Conference on Machne Learnng and Applcatons. Mam, USA: IEEE, Dec. 203, pp [8], Improvng Credt Card Fraud Detecton wth Calbrated Probabltes, n Proceedngs of the fourteenth SIAM Internatonal Conference on Data Mnng, Phladelpha, USA, 204, pp [9] R. J. Bolton, D. J. Hand, F. Provost, and L. Breman, Statstcal Fraud Detecton: A Revew, Statstcal Scence, vol. 7, no. 3, pp , [20] D. J. Hand, C. Whtrow, N. M. Adams, P. Juszczak, and D. J. Weston, Performance crtera for plastc card fraud detecton tools, Journal of the Operatonal Research Socety, vol. 59, no. 7, pp , May [2] A. Correa Bahnsen, D. Aouada, and B. Ottersten, Example- Dependent Cost-Senstve Decson Trees, Expert Systems wth Applcatons, vol. 42, no. 9, pp , 205. [22] F. Pedregosa, G. Varoquaux, A. Gramfort, V. Mchel, B. Thron, O. Grsel, M. Blondel, P. Prettenhofer, R. Wess, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay, Sckt-learn: Machne learnng n Python, Journal of Machne Learnng Research, vol. 2, pp , 20. [23] A. Correa Bahnsen, D. Aouada, and B. Ottersten, Example-Dependent Cost-Senstve Logstc Regresson for Credt Scorng, n 204 3th Internatonal Conference on Machne Learnng and Applcatons. Detrot, USA: IEEE, 204, pp APPENDIX The von Mses dstrbuton, also known as the perodc normal dstrbuton, s a dstrbuton of a wrapped normal dstrbuted varable across a crcle []. The von Mses dstrbuton of a set of examples D = {t, t 2,, t N } s defned as D vonmses (µ vm, /σ vm ), where µ vm and σ vm are the perodc mean and perodc standard devaton, respectvely, and are calculated as follows [2] µ vm (D) = 2 tan φ (, (9) ψ2 + φ 2 + φ) and where φ = ( ) σ vm (D) = ln ( N φ) 2 ( + N ψ) 2, (0) t j D sn(t j ) and ψ = cos(t j ). t j D

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis

The Development of Web Log Mining Based on Improve-K-Means Clustering Analysis The Development of Web Log Mnng Based on Improve-K-Means Clusterng Analyss TngZhong Wang * College of Informaton Technology, Luoyang Normal Unversty, Luoyang, 471022, Chna [email protected] Abstract.

More information

What is Candidate Sampling

What is Candidate Sampling What s Canddate Samplng Say we have a multclass or mult label problem where each tranng example ( x, T ) conssts of a context x a small (mult)set of target classes T out of a large unverse L of possble

More information

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol

CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK. Sample Stability Protocol CHOLESTEROL REFERENCE METHOD LABORATORY NETWORK Sample Stablty Protocol Background The Cholesterol Reference Method Laboratory Network (CRMLN) developed certfcaton protocols for total cholesterol, HDL

More information

Forecasting the Direction and Strength of Stock Market Movement

Forecasting the Direction and Strength of Stock Market Movement Forecastng the Drecton and Strength of Stock Market Movement Jngwe Chen Mng Chen Nan Ye [email protected] [email protected] [email protected] Abstract - Stock market s one of the most complcated systems

More information

The OC Curve of Attribute Acceptance Plans

The OC Curve of Attribute Acceptance Plans The OC Curve of Attrbute Acceptance Plans The Operatng Characterstc (OC) curve descrbes the probablty of acceptng a lot as a functon of the lot s qualty. Fgure 1 shows a typcal OC Curve. 10 8 6 4 1 3 4

More information

An Alternative Way to Measure Private Equity Performance

An Alternative Way to Measure Private Equity Performance An Alternatve Way to Measure Prvate Equty Performance Peter Todd Parlux Investment Technology LLC Summary Internal Rate of Return (IRR) s probably the most common way to measure the performance of prvate

More information

Bayesian Network Based Causal Relationship Identification and Funding Success Prediction in P2P Lending

Bayesian Network Based Causal Relationship Identification and Funding Success Prediction in P2P Lending Proceedngs of 2012 4th Internatonal Conference on Machne Learnng and Computng IPCSIT vol. 25 (2012) (2012) IACSIT Press, Sngapore Bayesan Network Based Causal Relatonshp Identfcaton and Fundng Success

More information

An Interest-Oriented Network Evolution Mechanism for Online Communities

An Interest-Oriented Network Evolution Mechanism for Online Communities An Interest-Orented Network Evoluton Mechansm for Onlne Communtes Cahong Sun and Xaopng Yang School of Informaton, Renmn Unversty of Chna, Bejng 100872, P.R. Chna {chsun,yang}@ruc.edu.cn Abstract. Onlne

More information

Traffic-light a stress test for life insurance provisions

Traffic-light a stress test for life insurance provisions MEMORANDUM Date 006-09-7 Authors Bengt von Bahr, Göran Ronge Traffc-lght a stress test for lfe nsurance provsons Fnansnspetonen P.O. Box 6750 SE-113 85 Stocholm [Sveavägen 167] Tel +46 8 787 80 00 Fax

More information

A Secure Password-Authenticated Key Agreement Using Smart Cards

A Secure Password-Authenticated Key Agreement Using Smart Cards A Secure Password-Authentcated Key Agreement Usng Smart Cards Ka Chan 1, Wen-Chung Kuo 2 and Jn-Chou Cheng 3 1 Department of Computer and Informaton Scence, R.O.C. Mltary Academy, Kaohsung 83059, Tawan,

More information

THE APPLICATION OF DATA MINING TECHNIQUES AND MULTIPLE CLASSIFIERS TO MARKETING DECISION

THE APPLICATION OF DATA MINING TECHNIQUES AND MULTIPLE CLASSIFIERS TO MARKETING DECISION Internatonal Journal of Electronc Busness Management, Vol. 3, No. 4, pp. 30-30 (2005) 30 THE APPLICATION OF DATA MINING TECHNIQUES AND MULTIPLE CLASSIFIERS TO MARKETING DECISION Yu-Mn Chang *, Yu-Cheh

More information

Single and multiple stage classifiers implementing logistic discrimination

Single and multiple stage classifiers implementing logistic discrimination Sngle and multple stage classfers mplementng logstc dscrmnaton Hélo Radke Bttencourt 1 Dens Alter de Olvera Moraes 2 Vctor Haertel 2 1 Pontfíca Unversdade Católca do Ro Grande do Sul - PUCRS Av. Ipranga,

More information

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ).

benefit is 2, paid if the policyholder dies within the year, and probability of death within the year is ). REVIEW OF RISK MANAGEMENT CONCEPTS LOSS DISTRIBUTIONS AND INSURANCE Loss and nsurance: When someone s subject to the rsk of ncurrng a fnancal loss, the loss s generally modeled usng a random varable or

More information

Can Auto Liability Insurance Purchases Signal Risk Attitude?

Can Auto Liability Insurance Purchases Signal Risk Attitude? Internatonal Journal of Busness and Economcs, 2011, Vol. 10, No. 2, 159-164 Can Auto Lablty Insurance Purchases Sgnal Rsk Atttude? Chu-Shu L Department of Internatonal Busness, Asa Unversty, Tawan Sheng-Chang

More information

How To Calculate The Accountng Perod Of Nequalty

How To Calculate The Accountng Perod Of Nequalty Inequalty and The Accountng Perod Quentn Wodon and Shlomo Ytzha World Ban and Hebrew Unversty September Abstract Income nequalty typcally declnes wth the length of tme taen nto account for measurement.

More information

A Hierarchical Anomaly Network Intrusion Detection System using Neural Network Classification

A Hierarchical Anomaly Network Intrusion Detection System using Neural Network Classification IDC IDC A Herarchcal Anomaly Network Intruson Detecton System usng Neural Network Classfcaton ZHENG ZHANG, JUN LI, C. N. MANIKOPOULOS, JAY JORGENSON and JOSE UCLES ECE Department, New Jersey Inst. of Tech.,

More information

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College Feature selecton for ntruson detecton Slobodan Petrovć NISlab, Gjøvk Unversty College Contents The feature selecton problem Intruson detecton Traffc features relevant for IDS The CFS measure The mrmr measure

More information

Study on Model of Risks Assessment of Standard Operation in Rural Power Network

Study on Model of Risks Assessment of Standard Operation in Rural Power Network Study on Model of Rsks Assessment of Standard Operaton n Rural Power Network Qngj L 1, Tao Yang 2 1 Qngj L, College of Informaton and Electrcal Engneerng, Shenyang Agrculture Unversty, Shenyang 110866,

More information

The Use of Analytics for Claim Fraud Detection Roosevelt C. Mosley, Jr., FCAS, MAAA Nick Kucera Pinnacle Actuarial Resources Inc.

The Use of Analytics for Claim Fraud Detection Roosevelt C. Mosley, Jr., FCAS, MAAA Nick Kucera Pinnacle Actuarial Resources Inc. Paper 1837-2014 The Use of Analytcs for Clam Fraud Detecton Roosevelt C. Mosley, Jr., FCAS, MAAA Nck Kucera Pnnacle Actuaral Resources Inc., Bloomngton, IL ABSTRACT As t has been wdely reported n the nsurance

More information

An Evaluation of the Extended Logistic, Simple Logistic, and Gompertz Models for Forecasting Short Lifecycle Products and Services

An Evaluation of the Extended Logistic, Simple Logistic, and Gompertz Models for Forecasting Short Lifecycle Products and Services An Evaluaton of the Extended Logstc, Smple Logstc, and Gompertz Models for Forecastng Short Lfecycle Products and Servces Charles V. Trappey a,1, Hsn-yng Wu b a Professor (Management Scence), Natonal Chao

More information

A Novel Methodology of Working Capital Management for Large. Public Constructions by Using Fuzzy S-curve Regression

A Novel Methodology of Working Capital Management for Large. Public Constructions by Using Fuzzy S-curve Regression Novel Methodology of Workng Captal Management for Large Publc Constructons by Usng Fuzzy S-curve Regresson Cheng-Wu Chen, Morrs H. L. Wang and Tng-Ya Hseh Department of Cvl Engneerng, Natonal Central Unversty,

More information

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek

THE DISTRIBUTION OF LOAN PORTFOLIO VALUE * Oldrich Alfons Vasicek HE DISRIBUION OF LOAN PORFOLIO VALUE * Oldrch Alfons Vascek he amount of captal necessary to support a portfolo of debt securtes depends on the probablty dstrbuton of the portfolo loss. Consder a portfolo

More information

Mining Feature Importance: Applying Evolutionary Algorithms within a Web-based Educational System

Mining Feature Importance: Applying Evolutionary Algorithms within a Web-based Educational System Mnng Feature Importance: Applyng Evolutonary Algorthms wthn a Web-based Educatonal System Behrouz MINAEI-BIDGOLI 1, and Gerd KORTEMEYER 2, and Wllam F. PUNCH 1 1 Genetc Algorthms Research and Applcatons

More information

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS

INVESTIGATION OF VEHICULAR USERS FAIRNESS IN CDMA-HDR NETWORKS 21 22 September 2007, BULGARIA 119 Proceedngs of the Internatonal Conference on Informaton Technologes (InfoTech-2007) 21 st 22 nd September 2007, Bulgara vol. 2 INVESTIGATION OF VEHICULAR USERS FAIRNESS

More information

Gender Classification for Real-Time Audience Analysis System

Gender Classification for Real-Time Audience Analysis System Gender Classfcaton for Real-Tme Audence Analyss System Vladmr Khryashchev, Lev Shmaglt, Andrey Shemyakov, Anton Lebedev Yaroslavl State Unversty Yaroslavl, Russa [email protected], [email protected], [email protected],

More information

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur

Module 2 LOSSLESS IMAGE COMPRESSION SYSTEMS. Version 2 ECE IIT, Kharagpur Module LOSSLESS IMAGE COMPRESSION SYSTEMS Lesson 3 Lossless Compresson: Huffman Codng Instructonal Objectves At the end of ths lesson, the students should be able to:. Defne and measure source entropy..

More information

7.5. Present Value of an Annuity. Investigate

7.5. Present Value of an Annuity. Investigate 7.5 Present Value of an Annuty Owen and Anna are approachng retrement and are puttng ther fnances n order. They have worked hard and nvested ther earnngs so that they now have a large amount of money on

More information

Statistical Methods to Develop Rating Models

Statistical Methods to Develop Rating Models Statstcal Methods to Develop Ratng Models [Evelyn Hayden and Danel Porath, Österrechsche Natonalbank and Unversty of Appled Scences at Manz] Source: The Basel II Rsk Parameters Estmaton, Valdaton, and

More information

Data Mining from the Information Systems: Performance Indicators at Masaryk University in Brno

Data Mining from the Information Systems: Performance Indicators at Masaryk University in Brno Data Mnng from the Informaton Systems: Performance Indcators at Masaryk Unversty n Brno Mkuláš Bek EUA Workshop Strasbourg, 1-2 December 2006 1 Locaton of Brno Brno EUA Workshop Strasbourg, 1-2 December

More information

Answer: A). There is a flatter IS curve in the high MPC economy. Original LM LM after increase in M. IS curve for low MPC economy

Answer: A). There is a flatter IS curve in the high MPC economy. Original LM LM after increase in M. IS curve for low MPC economy 4.02 Quz Solutons Fall 2004 Multple-Choce Questons (30/00 ponts) Please, crcle the correct answer for each of the followng 0 multple-choce questons. For each queston, only one of the answers s correct.

More information

Calculation of Sampling Weights

Calculation of Sampling Weights Perre Foy Statstcs Canada 4 Calculaton of Samplng Weghts 4.1 OVERVIEW The basc sample desgn used n TIMSS Populatons 1 and 2 was a two-stage stratfed cluster desgn. 1 The frst stage conssted of a sample

More information

STATISTICAL DATA ANALYSIS IN EXCEL

STATISTICAL DATA ANALYSIS IN EXCEL Mcroarray Center STATISTICAL DATA ANALYSIS IN EXCEL Lecture 6 Some Advanced Topcs Dr. Petr Nazarov 14-01-013 [email protected] Statstcal data analyss n Ecel. 6. Some advanced topcs Correcton for

More information

Latent Class Regression. Statistics for Psychosocial Research II: Structural Models December 4 and 6, 2006

Latent Class Regression. Statistics for Psychosocial Research II: Structural Models December 4 and 6, 2006 Latent Class Regresson Statstcs for Psychosocal Research II: Structural Models December 4 and 6, 2006 Latent Class Regresson (LCR) What s t and when do we use t? Recall the standard latent class model

More information

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting

Causal, Explanatory Forecasting. Analysis. Regression Analysis. Simple Linear Regression. Which is Independent? Forecasting Causal, Explanatory Forecastng Assumes cause-and-effect relatonshp between system nputs and ts output Forecastng wth Regresson Analyss Rchard S. Barr Inputs System Cause + Effect Relatonshp The job of

More information

Statistical Approach for Offline Handwritten Signature Verification

Statistical Approach for Offline Handwritten Signature Verification Journal of Computer Scence 4 (3): 181-185, 2008 ISSN 1549-3636 2008 Scence Publcatons Statstcal Approach for Offlne Handwrtten Sgnature Verfcaton 2 Debnath Bhattacharyya, 1 Samr Kumar Bandyopadhyay, 2

More information

ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING

ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING ANALYZING THE RELATIONSHIPS BETWEEN QUALITY, TIME, AND COST IN PROJECT MANAGEMENT DECISION MAKING Matthew J. Lberatore, Department of Management and Operatons, Vllanova Unversty, Vllanova, PA 19085, 610-519-4390,

More information

How To Understand The Results Of The German Meris Cloud And Water Vapour Product

How To Understand The Results Of The German Meris Cloud And Water Vapour Product Ttel: Project: Doc. No.: MERIS level 3 cloud and water vapour products MAPP MAPP-ATBD-ClWVL3 Issue: 1 Revson: 0 Date: 9.12.1998 Functon Name Organsaton Sgnature Date Author: Bennartz FUB Preusker FUB Schüller

More information

On-Line Fault Detection in Wind Turbine Transmission System using Adaptive Filter and Robust Statistical Features

On-Line Fault Detection in Wind Turbine Transmission System using Adaptive Filter and Robust Statistical Features On-Lne Fault Detecton n Wnd Turbne Transmsson System usng Adaptve Flter and Robust Statstcal Features Ruoyu L Remote Dagnostcs Center SKF USA Inc. 3443 N. Sam Houston Pkwy., Houston TX 77086 Emal: [email protected]

More information

Improving Credit Card Fraud Detection with Calibrated Probabilities

Improving Credit Card Fraud Detection with Calibrated Probabilities Improving Credit Card Fraud Detection with Calibrated Probabilities Alejandro Correa Bahnsen, Aleksandar Stojanovic, Djamila Aouada and Björn Ottersten Interdisciplinary Centre for Security, Reliability

More information

Calculating the high frequency transmission line parameters of power cables

Calculating the high frequency transmission line parameters of power cables < ' Calculatng the hgh frequency transmsson lne parameters of power cables Authors: Dr. John Dcknson, Laboratory Servces Manager, N 0 RW E B Communcatons Mr. Peter J. Ncholson, Project Assgnment Manager,

More information

CS 2750 Machine Learning. Lecture 3. Density estimation. CS 2750 Machine Learning. Announcements

CS 2750 Machine Learning. Lecture 3. Density estimation. CS 2750 Machine Learning. Announcements Lecture 3 Densty estmaton Mlos Hauskrecht [email protected] 5329 Sennott Square Next lecture: Matlab tutoral Announcements Rules for attendng the class: Regstered for credt Regstered for audt (only f there

More information

IMPACT ANALYSIS OF A CELLULAR PHONE

IMPACT ANALYSIS OF A CELLULAR PHONE 4 th ASA & μeta Internatonal Conference IMPACT AALYSIS OF A CELLULAR PHOE We Lu, 2 Hongy L Bejng FEAonlne Engneerng Co.,Ltd. Bejng, Chna ABSTRACT Drop test smulaton plays an mportant role n nvestgatng

More information

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network

Forecasting the Demand of Emergency Supplies: Based on the CBR Theory and BP Neural Network 700 Proceedngs of the 8th Internatonal Conference on Innovaton & Management Forecastng the Demand of Emergency Supples: Based on the CBR Theory and BP Neural Network Fu Deqang, Lu Yun, L Changbng School

More information

Mining Multiple Large Data Sources

Mining Multiple Large Data Sources The Internatonal Arab Journal of Informaton Technology, Vol. 7, No. 3, July 2 24 Mnng Multple Large Data Sources Anmesh Adhkar, Pralhad Ramachandrarao 2, Bhanu Prasad 3, and Jhml Adhkar 4 Department of

More information

Lecture 3: Annuity. Study annuities whose payments form a geometric progression or a arithmetic progression.

Lecture 3: Annuity. Study annuities whose payments form a geometric progression or a arithmetic progression. Lecture 3: Annuty Goals: Learn contnuous annuty and perpetuty. Study annutes whose payments form a geometrc progresson or a arthmetc progresson. Dscuss yeld rates. Introduce Amortzaton Suggested Textbook

More information

Analysis of Premium Liabilities for Australian Lines of Business

Analysis of Premium Liabilities for Australian Lines of Business Summary of Analyss of Premum Labltes for Australan Lnes of Busness Emly Tao Honours Research Paper, The Unversty of Melbourne Emly Tao Acknowledgements I am grateful to the Australan Prudental Regulaton

More information

A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION. Michael E. Kuhl Radhamés A. Tolentino-Peña

A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION. Michael E. Kuhl Radhamés A. Tolentino-Peña Proceedngs of the 2008 Wnter Smulaton Conference S. J. Mason, R. R. Hll, L. Mönch, O. Rose, T. Jefferson, J. W. Fowler eds. A DYNAMIC CRASHING METHOD FOR PROJECT MANAGEMENT USING SIMULATION-BASED OPTIMIZATION

More information

Predicting Software Development Project Outcomes *

Predicting Software Development Project Outcomes * Predctng Software Development Project Outcomes * Rosna Weber, Mchael Waller, June Verner, Wllam Evanco College of Informaton Scence & Technology, Drexel Unversty 3141 Chestnut Street Phladelpha, PA 19104

More information

Stochastic Protocol Modeling for Anomaly Based Network Intrusion Detection

Stochastic Protocol Modeling for Anomaly Based Network Intrusion Detection Stochastc Protocol Modelng for Anomaly Based Network Intruson Detecton Juan M. Estevez-Tapador, Pedro Garca-Teodoro, and Jesus E. Daz-Verdejo Department of Electroncs and Computer Technology Unversty of

More information

Design and Development of a Security Evaluation Platform Based on International Standards

Design and Development of a Security Evaluation Platform Based on International Standards Internatonal Journal of Informatcs Socety, VOL.5, NO.2 (203) 7-80 7 Desgn and Development of a Securty Evaluaton Platform Based on Internatonal Standards Yuj Takahash and Yoshm Teshgawara Graduate School

More information

Searching for Interacting Features for Spam Filtering

Searching for Interacting Features for Spam Filtering Searchng for Interactng Features for Spam Flterng Chuanlang Chen 1, Yun-Chao Gong 2, Rongfang Be 1,, and X. Z. Gao 3 1 Department of Computer Scence, Bejng Normal Unversty, Bejng 100875, Chna 2 Software

More information

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification

Logistic Regression. Lecture 4: More classifiers and classes. Logistic regression. Adaboost. Optimization. Multiple class classification Lecture 4: More classfers and classes C4B Machne Learnng Hlary 20 A. Zsserman Logstc regresson Loss functons revsted Adaboost Loss functons revsted Optmzaton Multple class classfcaton Logstc Regresson

More information

The Application of Fractional Brownian Motion in Option Pricing

The Application of Fractional Brownian Motion in Option Pricing Vol. 0, No. (05), pp. 73-8 http://dx.do.org/0.457/jmue.05.0..6 The Applcaton of Fractonal Brownan Moton n Opton Prcng Qng-xn Zhou School of Basc Scence,arbn Unversty of Commerce,arbn [email protected]

More information

Financial Mathemetics

Financial Mathemetics Fnancal Mathemetcs 15 Mathematcs Grade 12 Teacher Gude Fnancal Maths Seres Overvew In ths seres we am to show how Mathematcs can be used to support personal fnancal decsons. In ths seres we jon Tebogo,

More information

Statistical algorithms in Review Manager 5

Statistical algorithms in Review Manager 5 Statstcal algorthms n Reve Manager 5 Jonathan J Deeks and Julan PT Hggns on behalf of the Statstcal Methods Group of The Cochrane Collaboraton August 00 Data structure Consder a meta-analyss of k studes

More information

Effective Network Defense Strategies against Malicious Attacks with Various Defense Mechanisms under Quality of Service Constraints

Effective Network Defense Strategies against Malicious Attacks with Various Defense Mechanisms under Quality of Service Constraints Effectve Network Defense Strateges aganst Malcous Attacks wth Varous Defense Mechansms under Qualty of Servce Constrants Frank Yeong-Sung Ln Department of Informaton Natonal Tawan Unversty Tape, Tawan,

More information

Estimating Total Claim Size in the Auto Insurance Industry: a Comparison between Tweedie and Zero-Adjusted Inverse Gaussian Distribution

Estimating Total Claim Size in the Auto Insurance Industry: a Comparison between Tweedie and Zero-Adjusted Inverse Gaussian Distribution Avalable onlne at http:// BAR, Curtba, v. 8, n. 1, art. 3, pp. 37-47, Jan./Mar. 2011 Estmatng Total Clam Sze n the Auto Insurance Industry: a Comparson between Tweede and Zero-Adjusted Inverse Gaussan

More information

DEFINING %COMPLETE IN MICROSOFT PROJECT

DEFINING %COMPLETE IN MICROSOFT PROJECT CelersSystems DEFINING %COMPLETE IN MICROSOFT PROJECT PREPARED BY James E Aksel, PMP, PMI-SP, MVP For Addtonal Informaton about Earned Value Management Systems and reportng, please contact: CelersSystems,

More information

Transition Matrix Models of Consumer Credit Ratings

Transition Matrix Models of Consumer Credit Ratings Transton Matrx Models of Consumer Credt Ratngs Abstract Although the corporate credt rsk lterature has many studes modellng the change n the credt rsk of corporate bonds over tme, there s far less analyss

More information

Computer-assisted Auditing for High- Volume Medical Coding

Computer-assisted Auditing for High- Volume Medical Coding Computer-asssted Audtng for Hgh-Volume Medcal Codng Computer-asssted Audtng for Hgh- Volume Medcal Codng by Danel T. Henze, PhD; Peter Feller, MS; Jerry McCorkle, BA; and Mark Morsch, MS Abstract The volume

More information

Course outline. Financial Time Series Analysis. Overview. Data analysis. Predictive signal. Trading strategy

Course outline. Financial Time Series Analysis. Overview. Data analysis. Predictive signal. Trading strategy Fnancal Tme Seres Analyss Patrck McSharry [email protected] www.mcsharry.net Trnty Term 2014 Mathematcal Insttute Unversty of Oxford Course outlne 1. Data analyss, probablty, correlatons, vsualsaton

More information

Outsourcing inventory management decisions in healthcare: Models and application

Outsourcing inventory management decisions in healthcare: Models and application European Journal of Operatonal Research 154 (24) 271 29 O.R. Applcatons Outsourcng nventory management decsons n healthcare: Models and applcaton www.elsever.com/locate/dsw Lawrence Ncholson a, Asoo J.

More information

Forecasting and Stress Testing Credit Card Default using Dynamic Models

Forecasting and Stress Testing Credit Card Default using Dynamic Models Forecastng and Stress Testng Credt Card Default usng Dynamc Models Tony Bellott and Jonathan Crook Credt Research Centre Unversty of Ednburgh Busness School Verson 4.5 Abstract Typcally models of credt

More information

HOUSEHOLDS DEBT BURDEN: AN ANALYSIS BASED ON MICROECONOMIC DATA*

HOUSEHOLDS DEBT BURDEN: AN ANALYSIS BASED ON MICROECONOMIC DATA* HOUSEHOLDS DEBT BURDEN: AN ANALYSIS BASED ON MICROECONOMIC DATA* Luísa Farnha** 1. INTRODUCTION The rapd growth n Portuguese households ndebtedness n the past few years ncreased the concerns that debt

More information

IDENTIFICATION AND CORRECTION OF A COMMON ERROR IN GENERAL ANNUITY CALCULATIONS

IDENTIFICATION AND CORRECTION OF A COMMON ERROR IN GENERAL ANNUITY CALCULATIONS IDENTIFICATION AND CORRECTION OF A COMMON ERROR IN GENERAL ANNUITY CALCULATIONS Chrs Deeley* Last revsed: September 22, 200 * Chrs Deeley s a Senor Lecturer n the School of Accountng, Charles Sturt Unversty,

More information

How To Classfy Onlne Mesh Network Traffc Classfcaton And Onlna Wreless Mesh Network Traffic Onlnge Network

How To Classfy Onlne Mesh Network Traffc Classfcaton And Onlna Wreless Mesh Network Traffic Onlnge Network Journal of Computatonal Informaton Systems 7:5 (2011) 1524-1532 Avalable at http://www.jofcs.com Onlne Wreless Mesh Network Traffc Classfcaton usng Machne Learnng Chengje GU 1,, Shuny ZHANG 1, Xaozhen

More information

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic

Institute of Informatics, Faculty of Business and Management, Brno University of Technology,Czech Republic Lagrange Multplers as Quanttatve Indcators n Economcs Ivan Mezník Insttute of Informatcs, Faculty of Busness and Management, Brno Unversty of TechnologCzech Republc Abstract The quanttatve role of Lagrange

More information

Vision Mouse. Saurabh Sarkar a* University of Cincinnati, Cincinnati, USA ABSTRACT 1. INTRODUCTION

Vision Mouse. Saurabh Sarkar a* University of Cincinnati, Cincinnati, USA ABSTRACT 1. INTRODUCTION Vson Mouse Saurabh Sarkar a* a Unversty of Cncnnat, Cncnnat, USA ABSTRACT The report dscusses a vson based approach towards trackng of eyes and fngers. The report descrbes the process of locatng the possble

More information

Recurrence. 1 Definitions and main statements

Recurrence. 1 Definitions and main statements Recurrence 1 Defntons and man statements Let X n, n = 0, 1, 2,... be a MC wth the state space S = (1, 2,...), transton probabltes p j = P {X n+1 = j X n = }, and the transton matrx P = (p j ),j S def.

More information

A neuro-fuzzy collaborative filtering approach for Web recommendation. G. Castellano, A. M. Fanelli, and M. A. Torsello *

A neuro-fuzzy collaborative filtering approach for Web recommendation. G. Castellano, A. M. Fanelli, and M. A. Torsello * Internatonal Journal of Computatonal Scence 992-6669 (Prnt) 992-6677 (Onlne) Global Informaton Publsher 27, Vol., No., 27-39 A neuro-fuzzy collaboratve flterng approach for Web recommendaton G. Castellano,

More information

RequIn, a tool for fast web traffic inference

RequIn, a tool for fast web traffic inference RequIn, a tool for fast web traffc nference Olver aul, Jean Etenne Kba GET/INT, LOR Department 9 rue Charles Fourer 90 Evry, France [email protected], [email protected] Abstract As networked

More information

NEURO-FUZZY INFERENCE SYSTEM FOR E-COMMERCE WEBSITE EVALUATION

NEURO-FUZZY INFERENCE SYSTEM FOR E-COMMERCE WEBSITE EVALUATION NEURO-FUZZY INFERENE SYSTEM FOR E-OMMERE WEBSITE EVALUATION Huan Lu, School of Software, Harbn Unversty of Scence and Technology, Harbn, hna Faculty of Appled Mathematcs and omputer Scence, Belarusan State

More information

Data Visualization by Pairwise Distortion Minimization

Data Visualization by Pairwise Distortion Minimization Communcatons n Statstcs, Theory and Methods 34 (6), 005 Data Vsualzaton by Parwse Dstorton Mnmzaton By Marc Sobel, and Longn Jan Lateck* Department of Statstcs and Department of Computer and Informaton

More information

Data Broadcast on a Multi-System Heterogeneous Overlayed Wireless Network *

Data Broadcast on a Multi-System Heterogeneous Overlayed Wireless Network * JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 24, 819-840 (2008) Data Broadcast on a Mult-System Heterogeneous Overlayed Wreless Network * Department of Computer Scence Natonal Chao Tung Unversty Hsnchu,

More information

Classification of Network Traffic via Packet-Level Hidden Markov Models

Classification of Network Traffic via Packet-Level Hidden Markov Models Classfcaton of Network Traffc va Packet-Level Hdden Markov Models Alberto Danott, Walter de Donato, Antono Pescapè Department of Computer Scence and Systems Unversty of Naples Federco II {alberto, walter.dedonato,

More information

Meta-Analysis of Hazard Ratios

Meta-Analysis of Hazard Ratios NCSS Statstcal Softare Chapter 458 Meta-Analyss of Hazard Ratos Introducton Ths module performs a meta-analyss on a set of to-group, tme to event (survval), studes n hch some data may be censored. These

More information

A Probabilistic Theory of Coherence

A Probabilistic Theory of Coherence A Probablstc Theory of Coherence BRANDEN FITELSON. The Coherence Measure C Let E be a set of n propostons E,..., E n. We seek a probablstc measure C(E) of the degree of coherence of E. Intutvely, we want

More information

Trade Adjustment and Productivity in Large Crises. Online Appendix May 2013. Appendix A: Derivation of Equations for Productivity

Trade Adjustment and Productivity in Large Crises. Online Appendix May 2013. Appendix A: Derivation of Equations for Productivity Trade Adjustment Productvty n Large Crses Gta Gopnath Department of Economcs Harvard Unversty NBER Brent Neman Booth School of Busness Unversty of Chcago NBER Onlne Appendx May 2013 Appendx A: Dervaton

More information

CHAPTER 14 MORE ABOUT REGRESSION

CHAPTER 14 MORE ABOUT REGRESSION CHAPTER 14 MORE ABOUT REGRESSION We learned n Chapter 5 that often a straght lne descrbes the pattern of a relatonshp between two quanttatve varables. For nstance, n Example 5.1 we explored the relatonshp

More information

Time Value of Money Module

Time Value of Money Module Tme Value of Money Module O BJECTIVES After readng ths Module, you wll be able to: Understand smple nterest and compound nterest. 2 Compute and use the future value of a sngle sum. 3 Compute and use the

More information

Trivial lump sum R5.0

Trivial lump sum R5.0 Optons form Once you have flled n ths form, please return t wth your orgnal brth certfcate to: Premer PO Box 2067 Croydon CR90 9ND. Fll n ths form usng BLOCK CAPITALS and black nk. Mark all answers wth

More information

On the Optimal Control of a Cascade of Hydro-Electric Power Stations

On the Optimal Control of a Cascade of Hydro-Electric Power Stations On the Optmal Control of a Cascade of Hydro-Electrc Power Statons M.C.M. Guedes a, A.F. Rbero a, G.V. Smrnov b and S. Vlela c a Department of Mathematcs, School of Scences, Unversty of Porto, Portugal;

More information

Customer Lifetime Value Modeling and Its Use for Customer Retention Planning

Customer Lifetime Value Modeling and Its Use for Customer Retention Planning Customer Lfetme Value Modelng and Its Use for Customer Retenton Plannng Saharon Rosset Enat Neumann Ur Eck Nurt Vatnk Yzhak Idan Amdocs Ltd. 8 Hapnna St. Ra anana 43, Israel {saharonr, enatn, ureck, nurtv,

More information

Project Networks With Mixed-Time Constraints

Project Networks With Mixed-Time Constraints Project Networs Wth Mxed-Tme Constrants L Caccetta and B Wattananon Western Australan Centre of Excellence n Industral Optmsaton (WACEIO) Curtn Unversty of Technology GPO Box U1987 Perth Western Australa

More information

Gaining Insights to the Tea Industry of Sri Lanka using Data Mining

Gaining Insights to the Tea Industry of Sri Lanka using Data Mining Proceedngs of the Internatonal MultConference of Engneers and Computer Scentsts 2008 Vol I Ganng Insghts to the Tea Industry of Sr Lanka usng Data Mnng H.C. Fernando, W. M. R Tssera, and R. I. Athauda

More information

L10: Linear discriminants analysis

L10: Linear discriminants analysis L0: Lnear dscrmnants analyss Lnear dscrmnant analyss, two classes Lnear dscrmnant analyss, C classes LDA vs. PCA Lmtatons of LDA Varants of LDA Other dmensonalty reducton methods CSCE 666 Pattern Analyss

More information

Stress test for measuring insurance risks in non-life insurance

Stress test for measuring insurance risks in non-life insurance PROMEMORIA Datum June 01 Fnansnspektonen Författare Bengt von Bahr, Younes Elonq and Erk Elvers Stress test for measurng nsurance rsks n non-lfe nsurance Summary Ths memo descrbes stress testng of nsurance

More information

Improved SVM in Cloud Computing Information Mining

Improved SVM in Cloud Computing Information Mining Internatonal Journal of Grd Dstrbuton Computng Vol.8, No.1 (015), pp.33-40 http://dx.do.org/10.1457/jgdc.015.8.1.04 Improved n Cloud Computng Informaton Mnng Lvshuhong (ZhengDe polytechnc college JangSu

More information

A Performance Analysis of View Maintenance Techniques for Data Warehouses

A Performance Analysis of View Maintenance Techniques for Data Warehouses A Performance Analyss of Vew Mantenance Technques for Data Warehouses Xng Wang Dell Computer Corporaton Round Roc, Texas Le Gruenwald The nversty of Olahoma School of Computer Scence orman, OK 739 Guangtao

More information

An artificial Neural Network approach to monitor and diagnose multi-attribute quality control processes. S. T. A. Niaki*

An artificial Neural Network approach to monitor and diagnose multi-attribute quality control processes. S. T. A. Niaki* Journal of Industral Engneerng Internatonal July 008, Vol. 4, No. 7, 04 Islamc Azad Unversty, South Tehran Branch An artfcal Neural Network approach to montor and dagnose multattrbute qualty control processes

More information

Power-of-Two Policies for Single- Warehouse Multi-Retailer Inventory Systems with Order Frequency Discounts

Power-of-Two Policies for Single- Warehouse Multi-Retailer Inventory Systems with Order Frequency Discounts Power-of-wo Polces for Sngle- Warehouse Mult-Retaler Inventory Systems wth Order Frequency Dscounts José A. Ventura Pennsylvana State Unversty (USA) Yale. Herer echnon Israel Insttute of echnology (Israel)

More information