Iteratoal Joural of Computer Applcatos (0975 8887) Volume 119 No.5, Jue 2015 Multobjectve based Evet based Project Schedulg usg Optmzed Neural Network based ACO System Vdya Sagar Poam Research Scholar, CSE Departmet, Sr Krshadevaraya Uversty, Aathapuram, Ida N.Geethajal, PhD Assocate Professor & Head Dept. of Computer Scece & Techology, Sr Krshadevaraya Uversty, Aathapuram, Ida ABSTRACT I ay software project maagemet, developg thrd party software tools ad schedulg tasks are challegg ad mportat. Ay software developmet projects are flueced by a large umber of actvtes, whch ca greatly chage the project pla. These actvtes may form groups of correlated tasks or evet chas. Assessmet plag s a crucal challege software egeerg whose major goal s to schedule the persos to dfferet tasks such a way that the qualty of the software product s optmal ad the cost of the project should be mmum. I the tradtoal approach a evet depedet scheduler at coloy optmzato s appled o task schedulg.the ACO wll develop a optmzed pla, the form of matrx, from all the teratos. Ad from that pla the EBS(Evet Based Scheduler) wll develop schedule based o evets. ACO solves the problem of project schedulg, but t does ot cosder the updated task allocato matrx. The ACO s ot a satsfactory model to solve the problem of project schedulg. The tradtoal ACO system also dcates the problem of allocatg the detcal actvty for several umbers of employees varyg perods.i ths proposed work, a mproved ACO approach to optmal global search usg a eural approach was troduced to schedule multple tasks. A actvty wth specfed umber of tasks ad relevat resources ca be optmally scheduled usg mult-objectve approach. Whe a ucerta evet occurs the remag resources wll be effectvely calculated, also the remag tasks to complete. Ad aga a ew schedule wll be geerated accordg to t. A ehaced Etropy method ca be used to deote the level about how much threshold or formato has bee fgured out to the pheromoe trals ad subsequetly the heurstc parameter ca be mproved accordgly. Keywords Project maagemet,schedulg,task partto. 1. INTRODUCTION Software project professoals ackowledge the mportace of maagg ucertates. The teratve refemet approach, detecto ad research of possble egatve aspects ad utlzato of other best approaches ca scale dow ucertates ad help to brg the project as per the scope,stadared tme estmate ad prce [1,7]. Yet, software professoals have't bee always expert wth cotrollg approaches ad probablstc schedulg or choose as redudat overhead. Modelg the evet schedule wth ucertates the drecto of the plag part cotues to be crucal sce t allows the supervsor to estmate feasblty of the aalyze the project. Due to sgfcace ad challeges of project plag, there exsts a eed for better computer aded utltes for software project plag recet staces. To pla a software project, the project supervsor must assess the project operatos ad determe the project pla ad supply allocato. To boost more fascatg tools ad kds covetoal project plag approaches ought to be moreover exteded. I ths partcular paper, a good atttude for the task schedulg ad persoel allocato challege project plag utlzg a optmzed at coloy optmzato (ACO) algorthm s recommeded. Belleguez ad Ne ro [8], [9] utlzed a multskll schedulg method by extedg the ormal RCPSP method. The techque cosders both the ssues of huma resource allocato ad edeavor schedulg, ad takes the ablty profcecy of persoel ad supply multple tasks to cosderato. Tabu search space (TS) [10], brach ad certa [10], ad GA [11] are developed for the method. I each of the kds, there's a predcto that pre-empto s ot allowed. As outled, ths assumpto reduces the relablty of huma source allocato for software dutes. Exercse preempto applcato tasks s smply regarded a few research. I Chag s recet proposed system [3], he mproved hs covetoal schedulg atttude by troducg a 3D smlarty matrx llustrato, specfyg the workload actvty of each worker for each udertakg o every tme perod. Although ths llustrato s far more bedy, t makes the seek area house very huge ad suffers from the matter of task of workloads. I ths paper, a competet procedure for project schedulg ad huma source allocato challege s utlzed. It also cosders the usure pursuts hadlg. RCPSP- Resource Costraed Project Schedule Challege RCPSP asks a few crucal: Haded a mess of actos, a group of resources, alog wth a measuremet of effectvty, so whch s the deal system to assg the resources oto the actos a approach that the overall performace s maxmzed? Whch s the most effectve techque to assg the resources to the actos at specfc staces, so that every a part of the restrctos are satsfed ad aturally the deal goal measures are geerated? RCPSP ca smply be thought as follows: partcular actos that should be executed a multtude of resources wth whch to partcpate the actvtes, specfc restrctos, whch ought 21
Iteratoal Joural of Computer Applcatos (0975 8887) Volume 119 No.5, Jue 2015 to be satsfed a buch of targets wth whch eed to be fshed [1-5] I the paper[6] a optmal PSO approach for resourcecostraed project schedulg ssue was mplemeted. Ehacemets depeds o the smple partcle swarm optmzato corporate: the partcle s tated by heurstcrules to reforce the hgh qualtatve partcles; weght was selfadapted wth the terato of the approach to de-accelerate the velocty of partcles; crossover mechasm of flterg approach has bee utlzed to partcle swarm to permt the trade-off excellet trats betwee each partcle. The target of ths study was to lesse the overall software project spa tme. Computatoal software project crcumstaces of PSPLIB dsplay that ths mproved partcle swarm optmzato approach was valuable as cotrast wth other heurstc techques. I the YaLu lauched a ew fuzzy flterg algorthm for software project schedulg to overcome the challege wth supply costrats ad overall project perod [6,7]. Frstly, fuzzy set was used to symbolze the ucertaty of udertakg perod together wth the correspodg evaluato system of the fuzzy varety kow as the tegral worth method was lauched. Secod, three flterg operators have bee used to look for a approxmate shortest software project make spa. As a result, ths gaed kowledge of provdg aother metaheurstc procedure for fxg resource-costrat software project schedulg challeges wth a uclear edeavor perod. I the paper[5] Mohammad Am Rg, Shahrar Mohammad K. N. Toos mplemeted a brad ew method to supply costraed based software project schedulg ssue. A hybrd flterg algorthm such as costrat satsfacto, challege has bee used to fd source costraed based project schedulg. GA s edeavor was to locate the deal schedule. Ther method uses costrat satsfacto, challege wth the teto to overcome the preset cossteces pursuts prorty ad resource costrats. A full state costrat satsfacto, challege wth mmum coflct heurstc has bee used for fxg prorty coflcts ad a straghtforward teratve costrat satsfacto challege s used to resolve the orgal source coflcts. For mult-objectve RCPSP, Vaa ad Pho de Sousa [9] mplemeted ad reduce the mea weghted lateess of pursuts ad scale dow the sum of the volato of orgal source avalablty. Also, for the multdmesoal RCPSP, Dod & Elma [18] mplemeted the mxg of the elemets of tme, rate ad qualty o what they are kow as the Totally Optmzed Project System (TOPS). The TOPS may well be sad a follows: Determe the begg ad coclude tme of project actos (such as perod), the fabrc orderg ad stock suggestos, together wth the allocato of huma resources ad gear to those actos as a way to reduce the etre fee of the project or to optmze other selected measures of performace. They argued that ther bult- project schedulg challege could have less total rate tha depedetly cosderg ths goal a umarred goal challege. A essetal vestgate TOPS exhbts that the authors dd't cosder the effect of restrcted resources as clearly studed RCPSP. A cosderato of the tr- ature of project goal (tme, fee ad qualty) allows for a mult dmeso method to reachg the fal goal where resources are restrcted[10-14]. 2. PROPOSED APPROACH ACO [10] s a method based o the actvty of the task schedulg to fd a shortest path from a source task to the destato task. Ths method uses the actvty of the real-ats whle lookg for the food as optmal tasks. It was examed that the ats place a certa quatty of pheromoe ts path whle vstg from ts source place for the food. Cosstetly whle gog back, the ats exposed to follow the same drecto marked by the pheromoe place ad aga set the pheromoe ts path. I ths maer the ats follow the shortest-path are predcted to retur much earler ad thus ehace the quatty of pheromoe guaratee ts path at a qucker rate tha the ats focusg o a loger-path. Yet, the pheromoe s exposed to evaporato by a certa quatty at a cosstet rate over the certa terval ad cosequetly the paths vsted by the ats ofte, are oly kept as evdece by the pheromoe place, coversely the paths frequetly vsted by the ats are dropped due to the lack of pheromoe set o that path ad cosequetly the ew ats are plaed to follow the ofte used paths oly. Ad hece, most of the ats tatg ther jourey ca lear from the resources ad fo left by the formerly vstor ats ad are drected to follow the shortest-path structed by the pheromoe depost. I ACO approach, varous sythetc ats costruct solutos to the optmzato problem ad replace formato o the qualty of these solutos over a commucato pla. Italzato: Iput :J: Jobs R: Resources T: Tme Algorthm : ProjectIt(J,R,T) (JOB) R1 R2. R Fg 1. Resource Mappg T1 T2 T 22
Iteratoal Joural of Computer Applcatos (0975 8887) Volume 119 No.5, Jue 2015 Lst =getall(j);// Get all avalable jobs Lst β=getall(r);//get all avalable resources For each Job Do Lst φ=check the avalable resources (,β) Doe Map each job wth avalable resource as Matrx //check the mmum resource avalable tme Set the schedule_tme t = 0, j m rt For each avalable resource β ad avalable tme T Do Lst Temp1:=M{, β,t} //get all mmum resources wth mmum tme. Lst Temp2:= M{, β,t} Multple jobs resource costraed schedulg mathematcally derved as Optmal schedulg ca be acheved usg equato 1 mdurato =m{temp1,temp2} s.t Doe T t T -------------(1) j s j f Whe the mproved ACO method begs to operate, the capacty of formato o each sgle path balace wth oe aother, formato etropy(e) s optmal at ths tme, however a mprovemet of pheromoe o the path, the etropy value wll certaly be mmzed steadly. I case the etropy s ot maaged, the etropy wll ultmately scale dow to 0, usually, the pheromoe o oly oe path s optmal, ad the last soluto teds to be erroeous, thus carryg about the complete. I order to get over the easly occurred flaws for solvg challegg optmzato problems wth the basc algorthm, a proposed mproved aco based o optmzed squared formato etropy s metoed, applyg the heurstc parameter value selecto motored by squared formato etropy factor. 1. Basc Italze Step: Create tal feasble basc solutos ad fd out the approprate objectve fucto values; 2. Set prmary values of pheromoes ad other wth Temp1 ad Temp2. 3. Each at selects ts ow path to fd the optmal soluto. Each at decdes ts ext ode(job task) as per the task selecto process ad costructs the complete soluto matrx. 4. By employg the ftess fuctoalty, each at evaluates the optmal soluto ad also estmatet he durato of project,overwork ad cost for that project. 5. Classfy the all solutos usg Neural Network System ad Exame the all solutos ad decde the most sutable oe, revse the pheromoe value. 6. Perform repeatedly utl the codto s fulflled. Usually the termato codto s detfed by settg the umber of teratos. 7. Decde ad dsplay the best optmal soluto through whchdurato ad cost s less Calculate the etropy value of preset pheromoe trals ad the revse the heurstc parameters utl the codto s satsfed. Repeat Utl at k has fshed Ed I each tral a dscrete radom varable s selected from the pheromoe matrx. The ehaced etropy value of a radom varable s calculated as Et(X) plog(p ) where p deotes the probablty of occurrece of each trals the pheromoe matrx. Whe the probablty of the each tral s same the E wll be the optmal ad t s defed as MaxEt (X) 1/ r log(1/ r ) We recommed to utlse the squared etropy value as a dex to sgfy the degree about the level formato s gaed to the pheromoe trals ad the the heurstc parameter ca be updated approprately. MaxEt (X) 0.5* 1/ r log(1/ r ) MaxEt (X) 0.5* 1/ r [log(1) log(r )] MaxEt (X) 0.5* 1/ r [0 log(r )] MaxEt (X) 0.5* 1/ r [log(r )] 3. EXPERIMENTAL RESULTS I ths secto we preset the result of the computatos ad comparsos wth the best covetoal approaches. The approach s mplemeted java ad ru o a etbeas IDE. Well kow bechmark problem stace sets are used to evaluate the algorthm (PSPLIB, http://www.bwl.ukel.de/prod/psplb/dex.html). The sets j8, j10, j30, j60 cosst of more tha 500 problem staces are used our expermetal study. I our mplemetato, the eural learg coeffcet α s take as dyamcally ad the weghts are talzed at radom process. 23
Accuracy Iteratoal Joural of Computer Applcatos (0975 8887) Volume 119 No.5, Jue 2015 Fg 3. Task related Data Loadg Fg 5. Task Vs Days To complete Performace Aalyss Algorthm Accuracy Average Devatos GARCPSP 0.76 0.85 ACORCPSP 0.82 0.89 Proposed 0.96 0.98 1.2 1 Schedulg Accuracy 0.8 Accuracy 0.6 0.4 0.2 Average Devatos Fg 4. Ital Job Schedulg Before Optmzed ACO approach 0 Algorthms 24
Optmal Solutos Total Solutos Solved Iteratoal Joural of Computer Applcatos (0975 8887) Volume 119 No.5, Jue 2015 J60 PROJECT(600 Istaces) Algorthm Solved Total Optmalty(%) GARCPSP 345 600 0.575 ACORCPSP 367 600 0.611 Proposed 524 600 0.87 J60 JOB SCHEDULING 700 600 500 400 300 Solved 200 100 Total 0 ALGORITHMS J8 PROJECT(500 Istaces) Algorthm Solved Total Optmalty(%) GARCPSP 278 500 0.575 ACORCPSP 367 500 0.734 Proposed 467 500 0.93 J8 JOB SCHEDULING 600 500 400 Solved 300 Total 200 100 0 ALGORITHMS 4. CONCLUSION The ma objectve of ths paper s to solve the complcated plag problem, the secod oe method troduces a evet-based scheduler usg proposed approach. Both methods have lmtato durg the project plag ad allocato. Expermetal results show that the represetato scheme wth the EBS s effectve small target tasks, ad the mproved ACO algorthm maages to yeld better plas wth hgh statstc-t ad mea access tme ad more stable workload assgmets compared wth other exstg approaches. I ths proposed work, a mproved ACO approach wth optmal global search usg eural approach was troduced to schedule multple tasks. A actvty wth specfed umber of tasks ad relevat resources ca be optmally scheduled usg mult-objectve approach. Whe a ucerta evet occurs the remag resource wll be effectvely calculated, also the remag tasks to complete. Ad aga a ew schedule wll be geerated accordg to t. A ehaced Etropy method ca be used to deote the level about how much threshold or formato has bee fgured out to the pheromoe trals ad subsequetly the heurstc parameter ca be mproved accordgly. 5. REFERENCES [1] C. K. Chag, M. J. Chrstese, T. Zhag, (2001) Geetc algorthms for project maagemet, Aals of Software Egeerg, Vol. 11, pp107-139. [2] T. Hae & S. Nckel, (2005) A multobjectve evolutoary algorthm for schedulg ad specto plag software developmet project, Europea Joural of Operatoal Research, Vol. 167, pp 663-678. [3] J Leug, ( Ed), Hadbook of schedulg: algorthm models ad performace, CRC Press LLC; Florda. 2004 [4] E. Alba & J. F. Chcao, (2007) Software project maagemet wth GAs, Iformato Scece, Vol. 177, pp 2380-2401. [5] Mohammad Am Rg, Shahrar Mohammad K. N. Toos Fdg a Hybrd Geetc Algorthm- CostratSatsfacto Problem basedsoluto for ResourceCostraed Software project Schedulg Uversty of Techology, Idustral faculty, IT group Tehra, Ira, 2009 Iteratoal Coferece o Emergg Techologes. [6] Stso, J.P., Davs, E.W. ad Khumawala, B.M., "Multple Resourcecostraed Schedulg Usg Brach-ad-Boud", AIIE Trasactos, Vol. 10 No. 3, 1978, pp. 252 [10] Xggag Luo 1,2, Dgwe Wag 2, Jafu Tag 2, Ylu Tu 3ResourceCostraed Software project Schedulg Problem, Proceedgs of the 6th World Cogress o Itellget Cotrol ad Automato, Jue 21 23, 2006, Dala, Cha. [7] Ya Lu1,2,Sheg-L zharo2, X-Pg Zhag2, Guag- Qadu2, A GABased Approach for solvg fuzzy sftware project schedulg Proceedgs of the Sxth Iteratoal Coferece o Mache Learg ad Cyberetcs, Hog Kog, 19-22 August 2007. [8] O. Belleguez ad E. Ne ro, Methods for the Mult- Skll Project Schedulg Problem, Proc. Nth It l Workshop Project Maagemet ad Schedulg, 2004. 25
Iteratoal Joural of Computer Applcatos (0975 8887) Volume 119 No.5, Jue 2015 [9] P. Brucker, A. Drexl, R. Mohrg, K. Neuma, E. Pesch, Resource-Costraed Project Schedulg: Notato, Classfcato, Models ad Methods, Europea J. Operatoal Research, 1999. [10] O. Belleguez ad E. Ne ro, A Brach-ad-Boud Method for Solvg Mult-Skll Project Schedulg Problem, RAIRO- Operatos Research, 2007. [11] L.Ozdamar, A Geetc Algorthm Approach to a Geeral Category Project Schedulg Problem, IEEE Tras. Systems, Ma, ad Cyberetcs-Part C: Applcatos ad Rev., Feb 1999. [12] F. Kazem, R. Tavakkol-Moghaddam, Solvg a mult-objectve mult-mode Resource-costraed project schedulg problem wth partcle swarm optmzato, Iteratoal Joural of Academc Research, Vol. 3, pp. 103-110, 2011 [13] F. Ballest, R. Blaco, Theoretcal ad practcal fudametals for mult-objectve optmzato RCPSP, Joural of Computers ad Operato Research, Vol. 38, No. 1, pp. 51-62, 2011 [14] R. Akbar, V. Zegham, K. Zarat, Artfcal bee coloy for resource costraed project schedulg problem, Iteratoal Joural of Idustral Egeerg Computatos, Vol. 2, pp. 45-60, 2011 6. AUTHORS PROFILE P.Vdya Sagar obtaed hs M.C.A from Vsveswaraah Techologcal Uversty,Belgaum.The he obtaed hs M.Tech Computer Scece Ad Egeerg from Acharya Nagarjua Uversty,Gutur ad pursug PhD Computer Scece ad Techology from Sr Krshadevaraya Uversty,Aatapuram. He s a Professoal Member of ISCA. Hs specalzatos clude software egeerg ad software relablty, web servces ad etworkg. Dr.N.Geethajal receved her PhD Degree from Sr Krshaadevaraya Uversty. Adhra Pradesh, Ida she s workg as Head, Departmet of Computer Scece & Techology, Sr Krshaadevaraya Uversty. Adhra Pradesh, Ida. She s a Professoal Member ACM. Her research terest cludes Computer Networks, Cloud Computg, Software Egeerg, Programmg laguages ad Data Mg. IJCA TM : www.jcaole.org 26