Theta intelligent forecasting information system

The need effectively to integrate decision making tasks together with knowledge representation and inference procedures has caused recent research efforts towards the integration of decision support systems with knowledge-based techniques. Explores the potential benefits of such integration in the area of business forecasting. Describes the forecasting process and identifies its main functional elements. Some of these elements provide the requirements for an intelligent forecasting support system. Describes the architecture and the implementation of such a system, the theta intelligent forecasting information system (TIFIS) that that first-named author had developed during his dissertation. In TIFIS, besides the traditional components of a decisionsupport onformation system, four constituents are included that try to model the expertise required. The information system adopts an object-oriented approach to forecasting and exploits the forecasting engine of the theta model integrated with automated rule based adjustments and judgmental adjustments. Tests the forecasting accuracy of the information system on the M3-competition monthly data. to model the expertise required: a process expert, a data expert, a judgment expert and the theta expert an expert module that exploits the theta model as presented by Assimakopoulos and Nikolopoulos (1999, 2000), a model that performed particularly well in the M3-Competition (Makridakis and Hibon, 2000). The paper is structured as follows. The next section gives in brief the managerial implications of forecasting while in section an extended survey of business forecasting software is presented. In section 4 the proposed system is discussed in detail. The fifth section presents the evaluation of the system. The final section outlines the conclusions of the present work and gives directions for further research. 2. Forecasting: industrial and managerial implications Forecasting is one of the crucial factors so as to improve the performance of various industrial and managerial operations; it is important to firms because it can help ensure the effective use of resources (Klassen and Flores, 2001; Makridakis et al., 1998; Waddell and Sohal, 1994; Newbold and Bos, 1994). Nowadays, companies are in need of new forecast procedures, which will put them in a position of producing more accurate forecasts. Poor forecasts lead to inefficient capital management. In particular, the opportunities created by the use of a new but more accurate forecasting method are plenty and at the same time substantial for improving the functionality of the company (Zhao et al., 2002; Wacker and Sprague, 1998). More accurate forecasting on the company’s monthly sales will ensure better stock policy, more efficient warehouse management, better product distribution to the company’s branches and finally, minimization of the company’s risk in covering the market demands. Planning orders, requires precision so as to reduce the declinations from the final sales; accurate forecasts in ordering ensure cash management and cash flow optimization. Finally better sales’ forecast ensures better exchange policy for the transactions between the company and its clients. It is clear that managers can improve resource planning by understanding the limitations of forecasts. According to Wacker et al. (2002), these limitations are exemplified through several strategic forecasting paradoxes that managers should recognize. The paradoxes suggested by Wacker et al. (2002) are: the most important managerial decisions a company can make are based on the least accurate forecasts; the most useful forecast information for resource planning is the least accurate; and the organizations that need the most accurate forecast have the largest forecast error. By recognizing these paradoxes managers can devote their attention to improving the use and implementation of the forecast for better resource decisions. There are a number of forecasting models and software that are available to management and the choice of technique requires a number of considerations. If management believes that the future facing their company is predictable or fairly predictable, then statistical forecasting is a useful tool. On the other hand, if a company faces a very turbulent environment where the future is mostly unpredictable, then there is little point in attempting to utilize statistical techniques to forecast future (Lines, 1996). 3. Business forecasting software Seven large categories can be identified for business forecasting software (Nikolopoulos, 2002): 1 Commercial business forecasting packages. In the market today there is a great variety of specialized business forecasting packages. Their abilities and functionalities vary according to their price ($150-150,000). In the recent years several surveys have pointed out the pros and cons of such software (Yurkiewicz, 2000; KuÈ sters and Bell, 1999). According to the Yurkiewicz study the most popular are the following (in alphabetical order): Actuarial Forecast, AUTOBOX,B345 ProSeries Econometric System, Decision Pro, Decision Time and What If?, Forecast PRO, EViews 4, Forecast PRO Unlimited, Forecast PRO XE, ForecastX Engine, ForecastX Wizard, Fygir Demand Planning, GAUSS, Inventory Analyst Pro, LVCS, NCSS 2000, NeuralSIM, Peer Planner, Professional II Plus, PSI Planner, ROADMAP GENEVA, SCA Forecasting & Modeling Package, SmartForecasts for Windows, STATGRAPHICS, Statistica, STATLETS, Time Trends Forecast Warehouse. 2 Statistical packages. Statistical packages are often used for business forecasting. The most popular are the following four: [ 712] K. Nikolopoulos and V. Assimakopoulos Theta intelligent forecasting information system Industrial Management & Data Systems 103/9 [2003] 711-726 Minitab, SAS, SPSS, S-Plus (Makridakis et. al., 1998, pp. 578-9). 3 Spreadsheets. Spreadsheets (Microsoft Excel, Lotus 123, Quatro Pro) are widely used for simple business forecasting operations (the low cost along with user familiarity is a great advantage in this category). 4 Mathematical packages. Mathematical software (Matlab, MathCad, Mathematica) are also used for simple business forecasting operation, but their use requires advanced forecasting and mathematical skills. 5 ERPs. ERP systems (SAP, Oracle, Baan, e.t.a.) in the last decade have integrated business forecasting modules. The great disadvantage in this category is the prohibitive cost of the ERP for small businesses. 6 Academic forecasting information systems. Five academic forecasting information systems are widely referenced in the international bibliography: Flores/Pearce (Flores and Pearce, 2000; Vokurka et al., 1996), RBF (Adya et al., 2001; Adya, 2000; Adya et al., 2000; Collopy and Armstrong, 1992), AAM, (Melard and Pasteels, 2000), AutomatANN (Balkin and Ord, 2000), ARARMA (Meade, 2000). All these systems have participated in the M3-forecasting competition and produced promising results (Makridakis and Hibon, 2000) 7 e-Forecasting. We define e-forecasting as `̀ the ability of making forecasts by distance using the Internet’’. e-Forecasting stands for the most promising evolution of business forecasting software. A survey (Nikolopoulos, 2002) that took place recently indicated that there is very little research in the field of e-forecasting and there are not many forecasting services on the Internet. The survey was made under the basic assumption that a user wants and needs to have forecasts for his data over the Internet only by using a Web