A performance boosting transition in predictive modelling for customer acquisition

Improving business performance requires accurate prediction and decision-making, which are informed by historical data insights such as user attention, action, and profit from the products. These insights are generated using statistical techniques that forecast business needs and analyse data to make informed business decisions. Using historical data, a predictive model is built using a statistical approach, commonly employing methods such as regression, time series and cluster analysis. Machine Learning (ML) algorithms enhance this process by automating tasks and uncovering insights. Traditional analysis methods struggle with the complex pattern and dynamic nature of the data, leading to difficulties in interpreting relationships. To address this issue, the Performance Boost - Predictive Model (PB-PM) is proposed, which comprises two levels: level 1 employs Linear Regression (LR) and level 2 employs Ridge regression, Lasso regression, Random Forest (RF) regression and Extreme Gradient Boost (XGBoost) regression. The performance of the PB-PM model is evaluated using Mean Absolute Error (MAE) and R-squared score (R²). The analysis result indicates XGBoost regression algorithm yields the highest R² and lowest MAE for the proposed PB-PM, respectively 0.23 and 0.990.