Classification of transaction anomalies in the fuel industry using data mining methods

Data mining applications in the fuel industry are developing day by day and the applications are becoming widespread. With the methods used in the fuel industry and various analyses, important issues such as fuel theft, leakage, quantity excesses at the time of filling, and overflow after overfilling are monitored and actions are taken. In this study, different data mining classification methods were applied using the data of an oil company operating in Turkey between 2022-2023 in order to measure the accuracy and performance of the analyzes by classifying the analyzed and categorized data with different data mining classification methods in line with the 4 important categories determined in this study. Before the application, the data was made suitable for analysis. Variables that had no effect in the analysis were removed. Since opening, closing, filling, difference, sales, SEL value, decrease amount and category are the determining factors in the analysis, these data were used in the analysis. RAPIDMINER program was used during the implementation phase. Classifications were made with the k nearest neighbor algorithm, Random Forest Algorithm, Gradient Boosted Algorithm, ADABOOST Algorithm and Decision Tree (J48) Algorithm, which are among the data mining classification methods, and the success of the models was measured with various success criteria. It has been observed that the most successful classification method is the Decision Tree (J48) Algorithm