Predicting Customer Behavior from Transaction Data Using Machine Learning Classification Techniques

Knowing how customers act is important for increasing customer involvement, keeping customers, and making the business successful in today’s competitive market. Utilising transaction data, this study paper describes a machine learning-based model that can guess how customers will act in the future. Based on customers’ buying habits and personal data, the study uses several classification methods, such as Decision Trees, Random Forest, Support Vector Machines (SVM), and Logistic Regression, to put them into behavioural groups. We used two years of purchasing and demographic data from a store, which includes things like customer ID, transaction date, items bought, transaction value, and payment method. Steps in preprocessing include storing category factors, dealing with missing values, and normalising transaction amounts. Feature engineering adds new factors like Recency, Frequency, and Monetary Value, which are significant in predicting how customers will act. Fivefold cross-validation was used to train and test the models, and their success was judged by their F1-score, accuracy, precision, and recall. Overall, Random Forest was the most accurate algorithm (89.3%), closely followed by SVM (87.5%). Because it was easier to understand, logistic regression was not as accurate. According to the study of feature importance, the most important factors that predicted customer behaviour were the number of transactions, how recently they happened, and the average value of those transactions. The findings of this study show that machine learning classification methods can accurately guess how customers will act. This gives businesses useful information for improving their marketing strategies and keeping customers interested. For more accurate results and more business uses, future research could look into adding more data sources, real-time prediction models, and deep learning methods to the model.