An LSTMRM-Based E-Commerce Customer Churn Prediction System in Iot with Cloud Environment

The customer churn prediction is crucial for e-commerce business as it enables them to develop effective customer retention strategies and execute successful marketing operations. The existing works focus on the correlation rather than the causal factors, so it is difficult to inform the effective interventions. This paper introduces the LSTMRM approach for the effective customer churn prediction. First, the dataset of the customer behavior is collected. Next, preprocessing involves the imputation of missing values with sparsity handling using the KNTSN approach; then, de-duplication and data type conversion are done. After that, from the preprocessed results, dynamic customer segmentation is done by KH-PKC, and simultaneously feature extraction occurs. Then, the customer churn is defined by the CPH model; after that, CFGL is utilized based on the CPH output and the extracted features. Next, the trend and seasonal pattern are recognized by using the DWC-CHT. After that, DTW visualization is done from the results of the trend and seasonal pattern, and feature extraction takes place. Next, from the extracted features of DTW visualization, the churn is predicted by the LSTMRM approach. Finally, the predicted outcome from the LSTMRM approach and the churn-defined outcome from the CFGL model are stored in a cloud server. As per the experimental analysis, the proposed model attained 98.95% accuracy.