Classification of stocks with multi-criteria decision making methods and machine learning techniques

The purpose of this study is to put an effective hybrid model for stock management by using Multi Criteria Decision Making (MCDM) methods and Machine Learning (ML) methods together. For this purpose, Gray Relational Analysis (GRA) and Reference Ideal Method (RIM), which are among MCDM methods, and Artificial Neural Networks (ANN), Random Forest (RF), Naive Bayes (NB) and K Nearest Neighborhood (KNN) methods, among the machine learning methods, were used. While RIM calculates with the ideal point or the ideal range, the GIA uses the ideal point. In this respect, they differ from other MCDM methods. Within the scope of the study, 1250 stock items of an automobile spare parts retailer enterprise were classified. The criteria in the literature were examined and it was determined that the criteria for quality, output quantity, price, profitability, availability and market share are important for the classification of automobile spare parts. The weights of the criteria were determined by the Max100 method. By using the GIA and RIM methods, the stocks' rankings of importance were established by using the values that the stocks received according to the criteria. In order to estimate the classes of inventory items classified by ABC analysis, application was made with ma hine learning methods. Ideal range or ideal values are used in quality, output quantity, price, profitability and market share criteria. It has been concluded that the classification results obtained by RIM, which calculates using the ideal range, give better results than the GRA method, resulting in stock items that are important for the business. It has been demonstrated that machine learning methods perform well in estimating classes created by GRA and RIM. By making an application in a spare parts retailer business, important criteria for spare parts were determined and contribution was made to the literature. By using balanced and proportional data sets, estimation of stock items was made with Machine Learning methods. It has been concluded that the most successful method in estimating the classes of stock items is the Random Forest method.