Customer purchase prediction in electronic markets from clickstream data using the Oracle meta-classifier

Electronic commerce has brought about a fundamental change in marketing atmosphere. Marketers strive to gain a competitive advantage by providing exciting electronic platforms to attract visitors. Suppliers must comprehend and fulfill their buyers’ demands to increase their income and profit. Analyzing users’ behavior and predicting their purchase attitudes significantly influences marketing strategies. A comprehensive prediction of customers’ purchase intention is crucial to developing good marketing strategies, which may trigger much greater purchase amounts. Marketing analysts utilize various data mining approaches to identify consumer traits and intentions, helping marketing analyzers distinguish complex consumption patterns. This paper aims to recommend an ensemble meta-classifier algorithm for predicting customer purchase intentions in electronic markets using their clickstream data. To achieve this, we performed the RFECV procedure to select the most practical features, thus achieving the highest accuracy. Additionally, we applied hyperparameter optimization and tuning to select the best parameters for prediction. Furthermore, we chose to use Oracle as a static ensemble meta-classifier method. The results demonstrate that using the Oracle algorithm in combination with supervised classifiers and RFECV feature selection achieves higher precision and a lower error rate compared to conventional classifiers.