DOI: 10.14355/ijams.2014.0301.11 The Integrated Inventory Management with Forecast System Noor Ajian Mohd Lair *1, Chin Chong Ng 2, Abdullah Mohd Tahir, Rachel Fran Mansa, Kenneth Teo Tze Kin 1 School of Engineering and Information Technology, Universiti Malaysia Sabah, Jalan UMS, 88400 Kota Kinabalu, Sabah, Malaysia *1 nrajian@ums.edu.my; 2 c_chong89@hotmail.com Abstract Inventory Management System is very important for driving a company into better achievement. The main objective of inventory management is to keep the inventory level as low as possible and meet customers demand. This project centers on development of the computerized integrated inventory management and forecast system for the Guan Lee Sdn Bhd. In this project, the exponential smoothing is selected to predict demands as inputs to control the computerised inventory management system. The integrated system was written using the Visual Basic 2008. This integrated inventory management and forecast system has the ability to forecast while effectively control on inventory level with six specific features of alert, creation, inventory, transfer, search and reports. Performance of the system was analyzed with three types of forecast value (actual and adjusted forecast values). From the results, the actual forecast values tend to move toward the ideal one. Thus, the forecast system is proved to be reliable and accurate. The suggested improvements are auto recording of historical data, attachment of picture to each item, and sending notification through text message. Keywords Inventory Management System; Forecast Technique; Exponential Smoothing; Mean Absolute Deviation Introduction Inventories are materials and supplies that a business or institution carries for either sale or providing inputs or supplies to production process (Wildm, 2002). Improving the inventory management system enables a company to keep track on their inventory level consistently while supplies customers needs and maintains their inventory level as low as possible with minimum cost. Selection and utilization of an appropriate method and software for a computerized inventory management system is very important for company in order to be more efficient. This project centered on improving the Guan Lee Sdn Bhd inventory management strategies. The company controls the inventory manually without any inventory management system. Few major problems were incurred due to mis managed inventory. The time for stocks being used up and re reorder, is difficult to be determined by the company. The annual demand for the company is uncertain and hard to be predicted precisely, as it strongly depends on season, marketing, management and etc. The monthly quantity ordered by the company is imprecise, and always causes excessive or insufficient stocks. Therefore, the objective of this project is to develop an efficient computerized inventory management system that helps the company to control and manage the inventories through efficient forecasting system. Review of Literature In inventory management, various techniques have been used to manage the stock. For instance, the step function can be used to represent many real life situations in which the storage items can be classified into different ranges, each with its distinctive unit holding cost (Alfares, 2007). Alfares (2007) introduced two types of discontinuous step functions to represent these holding costs which are retroactive increase and incremental increase holding costs. For retroactive increase, a uniform holding cost that depends on the length of storage is used. For incremental increase, higher storage cost rates is applied to storage in later periods. By using these step functions, the respective total inventory cost (TIC) coupled with the ordering cost is then developed for further calculation in further research. The Croston s method used to predict the inventories with intermittent demand is an adaptation of the exponential smoothing proposed by Croston in 1972, involving separate simple exponential smoothing (SES) forecasts on the demand size and the demand interval (Synder, 2002). Later, another improvement on the 50
International Journal of Advances in Management Science (IJ AMS) Volume 3 Issue 1, February 2014 www.ij ams.org Croston s method was introduced by Synder (2002), called the adaptive variance version (AVAR). Synder proposed modifications to overcome certain implementation difficulties in forecasting slow and fast moving inventories. In the paper, Synder introduced variance instead of mean absolute deviation (MAD) for measuring variability in a time series, and a second smoothing parameter β to define how the variability changes over time, for Croston s method. Another new method proposed by Teunter et al. (2011), is called TSB (derived from authers name), which is a modification of Croston s method as well. In that modification, exponential smoothing was utilized to update the demand probability instead of the demand interval. The estimate of the probability of occurrence is updated at eavh time period. The estimate of the demand size is updated at the end of periods with positive demand. Then, two different smoothing constants were applied because the demand probability is updated more often than the demand size. Thus, the product of the estimates for demand size and demand probability provides the forecast of the demand per period. Periodic review system (R,S) is another type of inventory management policy used to deal with highly variable and irregular demand, where R stands for the review period while S is the base stock. At each review instance, the order quantity for any item is S IP, where IP is the inventory position of that item, namely the stock is either physically available or has been previously ordered but not yet received. In 2010, Nenes et al.(2010) adopted and implemented the periodic review system (R,S) to solve the problem of managing the inventories of thousands of different items, supplied by more than 20 European and Asian manufacturers and sold to a large number of differenttype customers. The lead time for every supplier is unlike to each other s. Thus, those researchers use this method as the review period R can be used regarding to all different suppliers. In addition, the bootstrap is a method that creates pseudo data by sampling with replacement from the individual observations (Willemain et al., 2004). In the problem of forecasting lead time demand, Willemain et al. (2004) adopted this method, and developed a modified bootstrap in response to three difficult features of intermittent demand, which are autocorrelation, frequent repeated value, and relatively short period. In their research, Markov model was used to generate a sequence of zero/nonzero value over forecast horizon. Summing the forecast over each period of the lead time gives one forecast of lead time demand (LTD). Thus, the process was repeated until they have 1000 bootstrap forecasts estimating the entire distribution of LTD. A method called fuzzy set theory was also applied in inventory problem, which can be found in the field of artificial intelligence either. Fuzzy set theory is concerned with the rules for computing the combined possibilities over expressions that contain fuzzy variable (Luger, 2005). For instance, a model constructed by Kao and Hsu (2002) for the case of fuzzy demand, was adopted as fuzzy number that was described by a membership function. After that, the total cost was computed from the membership function in term of fuzzy numbers for three different cases. As fuzzy number can be ranked, then Yager s method was applied for ranking the fuzzy numbers. At the end, a quantity with the smallest fuzzy cost (optimal quantity) was calculated. Decomposition procedures are used in time series to describe the trend and seasonal factors in a time series. By using decomposition procedures, seasonal component of time series, which influences the original time series, can be removed. For instances, Gardner Jr. and Diaz Saiz (2002) conducted their research coupled with an additive decomposition procedure for seasonal adjustment of inventory demand series at a large US auto parts distributor, BPX. In adjustment of seasonal series, first of all, the nature of demand series was identified on whether it is seasonal or not by comparing the variance of original series with the seasonally adjusted series, and then additive adjustment was applied instead of multiplicative adjustment on the series. This research attempts to integrate the inventory management and forecast system in contolling and managing the inventory for a company. Specifically, the exponential smoothing forecast technique will be integrated into a computerised inventory management system. The Case Study The Guan Lee Sdn Bhd commenced business in 1998 as a store selling daily necessities at Bayan Lepas, Penang. The inventory of the store is practically well managed as the volume of the goods is fair enough to be arranged systematically. After seven years, the company expanded their core business to sell bicycle. However, over the years, the business started growing 51
as a major supplier of bicycle of that area even further. At that time, the volume of inventory was very high and messy, and tracking the amount of each good manually was no longer feasible. In 2007, the company moved the business of mat from the first store to another new store, with the focus on providing mat only. Currently, the Guan Lee Sdn Bhd owns 3 stores and a storehouse at Bayan Lepas, Penang. The storehouse is fully occupied with the inventory for the 3 stores with very limited space. The entire storehouse is managed solely by the owner, without any computerised inventory management system. Currently, the Guan Lee Sdn Bhd is run by seven peoples consisting of one director, three supervisors and three workers. The storehouse consists of four sections, three sections at first floor, which are A, B, and C from front entrance to rear entrance, and the last section at second floor, which is the D section. In order to specify the location of the items within the section, each section is divided again into five subsections, rangeing from 1 to 5 subsections. The Integrated Inventory Management and Forecast System The Integrated Inventory Management and Forecast System consists of two distinct system; the inventory management system and the demand forecast system. Basically, the inventory management system offers 6 features (alert, transfer, creation, inventory, search, reports) for the user to manipulate the inventory of the storehouse. With the system, user is able to store the quantity of each item inside the storehouse with complete information such as location, category and etc. Plus, user will be notified by the system itself as critical circumference occurs such as extremely low inventory level. The demand forecast system anticipates the future demands for the company. The system uses the exponential smoothing technique to predict the future demands which are then used as an inputs for the quantity to be ordered by the company. The general structure of the forecast system is shown in Fig. 1. The formula used in the exponential smoothing technique is shown below: Where, Forecast for period Forecast for period Smoothing constant Actual demand or sales for period FIG. 1 GENERAL STRUCTURE OF THE FORECAST SYSTEM Performance Analysis of the Forecast System The performance of the system was analysed on 3 selected items with its historical monthly demands (12 months), to determine whether the forecast values generated with minimum mean absolute deviation (MAD) value are reliable or not. A simulation of the forecasting function of the system was conducted, within the 12 months of year 2010 and 2011, for analysis. In addition, 3 types of forecast values (actual, maximum and minimum adjusted forecast values) were considered together with the actual demand. A simulation of the forecasting function of the system was conducted by entering the actual demand one by one to obtain the individual forecast value (called the actual forecast) for each month before another. Each actual demand entered was computed 46 times to obtain a smoothing constant with minimum MAD, thus the constant for every month might be different from each other. Apart from the actual forecast values, the adjusted forecast values were also obtained by entering all the actual demands in one time. Once all the actual demands were entered, the entire forecast values (adjusted forecast values) were then computed, by adjusting the forecast values in the 12 months until the one with minimum MAD (called the ideal forecast), and another with maximum MAD (called the undesired forecast). Table 1 shows the history data along with the forecast values and individual error for 2011. The value of the initial forecast, which is 46 in January is computed from the exponential smoothing equation. The smoothing constant for the forecast technique used is 0.42. For bicycle with size of 26 inches, according to Fig. 2, both MAD of year 2010 and 2011 show a parabolic pattern with maximum point toward left side, but they converge toward right side, with an approximately same turning point. In year 2010, MAD rose to the maximum point at first, and decreased to the turning 52
International Journal of Advances in Management Science (IJ AMS) Volume 3 Issue 1, February 2014 www.ij ams.org point, but continued to decrease to a point where MAD is 0.5 at smoothing constant of 0.5. However, in year 2011, same behaviour as year 2010 at the beginning, but there is a minimum MAD at that turning point, which is 0.42 at smoothing constant of 0.42. TABLE 1 FORECAST VALUES AND INDIVIDUAL ERRORS OF 26 INCHES BICYCLE (2011) forecast values have fine smoothing effect and sensitivity. However, the MAD of the undesired forecast values is lower than the actual one, which might due to the disturbance of the actual forecast pattern in September. In addition, the actual forecast values for 2011 tend to move toward the ideal one as well. From the results, the overall actual forecast values tend to move toward the ideal one which is significantly closed to either, even the MAD is greater than the undesired one. From that, this result proved that the forecast system is reliable and accurate for forecasting. Apart from that, the pattern of the forecast values is not clearly observed, as the range of period is limited within 12 months only. Thus, a wider range of period should be considered for further analysis, such as weekly or even daily with fast moving item. Lastly, from this analysis, the results indicated that a seasonal demand is much more compatible with the exponential smoothing model, as the demand pattern has a trend. Conclusions FIG. 2 MAD VERSUS SMOOTHING CONSTANT OF 26 INCHES BICYCLE The developed integrated inventory management and forecast system offers fast response to current inventory on hand at any time. User is able to immediately responses to customer on whether a particular requested item is available or not. Besides that, forecasting, summarizing data and analysis are easily performed with aid of the system. Data can be fully accessable for user to gather all information whenever needed for any purpose such as analysis or forecasting. In addition, user can identify which product the best seller from the data. In addition, the developed system is able to function well as the actual forecast values tend to move toward the ideal one ; as well to generate a smoothing constant with the minimum MAD within the smoothing constant ranging from 0.45 to 0.50, for every actual demand entered by the user. The forecast system is reliable and accurate according to the results. FIG. 3 COMPARISONS BETWEEN DEMANDS AND 3 TYPES OF FORECAST VALUES FOR THE 26 INCHES BICYCLE (2011) Fig. 2 shows the graph of MAD versus the smoothing constant for the 26 inches bicycle. This graph is plotted according to the data generated from the inventory system internally as well. Fig. 3 shows the comparisons between demands and three types of forecast values for the 26 inches bicycle, in the year 2010 and 2011 respectively. According to Fig. 3, all the However, the developed system does have limitations; one of which discovered from this system, is lack of visualized effect on a particular item, or being difficult to be identified from item code or name. Thus, a picture should be attached to every item to improve effectiveness of the system. In addition, the system is still lack of automated feature, and user has to intentionally manipulate the entire system without any supportive suggestion from the system. Thus, 53
features such as making order automatically, or sending notification through text message and etc will be useful. ACKNOWLEDGMENT The authors would like to acknowledge the Malaysian Ministry of Higher Learning for the FRGS grant awarded for this project. REFERENCES Alfares H. K. Inventory model with stock level dependent demand rate and variable holding cost. International Journal of Production Economics, 108 (2007): 259 265. Gardner Jr, E. S., and Diaz Saiz, J. Seasonal adjustment of inventory demand series: a case study. International Journal of Forecasting, 18 (2002): 117 123. Kao, C., and Hsu, W.K. A single period inventory model with fuzzy demand. Computer and Mathematics with Applications, 43 (2002): 841 848. Luger, G.F. Artificial Intelligence. 5 th Ed. London: Pearson Education, 2005. Nenes, G., Panagiotidou, S., and Tagaras, G. Inventory management of multiple items with irregular demand: a case study. European Journal of Operational Research, 205 (2010): 313 324. Synder, R. Forecasting sales of slow and fast moving inventories. European Journal of Operational Research, 140 (2002): 684 699. Teunter, R. H., Syntetos, A. A., and Babai, M. Z. Inttermittent demand: linking forecasting to inventory obsolescence. European Journal of Operational Research, 214 (2011): 606 615. Wildm T. Best Practice in Inventory Management. 2nd Ed. Burlington, New York: Elsivier Butterworth Heinemann, 2002. Willemain, T.R., Smart, C.N., and Schwarz, H. F. A new approach inttermittent demand for service parts inventories. Journal of Forecasting, 20 (2004): 375 387. Dr Noor Ajian Mohd Lair is a senior lecturer in the Mechanical Engineering Program of Universiti Malaysia Sabah (UMS). She graduated with a Bachelor of Science in Industrial Engineering from University of Missouri Columbia (MU), USA in 1995 and Master of Mechanical Engineering from Universiti Teknology Malaysia, Malaysia in 2003. She received her Doctor of Philosophy (PhD) degree from University of South Australia (UniSA), Australia in 2009. Dr Noor Ajian areas of expertise include Supply Chain Management, Production Planing and Contol, Plant Optimisation and Operation Research using Simulation Modeling and Artificial Intelligent. 54