Precision Identification of Marketing Customers using Long Short-Term Memory with Multi-Layer Perceptron

In recent years, understanding customer behavior and achieving precise marketing have become crucial for enterprises, as this can provide valuable insights into customer preferences and the marketing decision-making process. It allows businesses to modify their strategies of marketing along with contributions that increase loyalty and satisfaction of client. However, the existing models struggles with varied customers behaviors in real-time data. Hence to address these challenges, Long Short-Term Memory with Multi-Layer Perceptron (LSTM-MLP) framework is proposed in this research. The proposed LSTM-MLP framework utilizes tags of Radio-Frequency Identification (RFID) to track the movement of product and gathers data on actions of customer utilizing tags Receive Signal Strength (RSS). Further, K-Nearest Neighbor (KNN) imputation is utilized for handling missing data in preprocessing. Further, isolation Forest (iForest) is used to eliminate outliers by identifying and eliminating anomalous data points. Consequently, LSTM is introduced to extract features by capturing patterns and temporal dependencies in sequential data. After that, the Adaptive Synthetic Sampling (ADASYN) is utilized to balance the data and it is followed by employing MLP for classifying customer shopping behaviors. From the results, the proposed LSTM-MLP framework outperformed existing MLP model in terms of accuracy (98.69%), precision (99.06%), recall (99.56%), F1 score (98.65%) and specificity (99.68%) respectively. The proposed customer behavior recognition model can provide more powerful decision-making basis for precision marketing.