Predictive modelling of omni-channel customer behavior using big data analytics for retail marketing

This study aims to develop an integrated, data-driven framework to predict omni-channel customer behavior by leveraging both structured data (e.g., RFM metrics) and unstructured data (e.g., sentiment from customer reviews). The goal is to understand how behavioral and emotional indicators influence conversion and loyalty across digital and physical retail channels. The research adopts a quantitative approach combining classical statistics and machine learning models. Structured and unstructured data were merged using Python-based tools. Sentiment was extracted via VADER and TextBlob, while engagement metrics were reduced using Principal Component Analysis (PCA) into a unified index. Predictive modelling was performed using Logistic Regression and Random Forest classifiers. Statistical testing (t-tests, ANOVA) and interaction term analysis assessed the moderating effect of channel type. The framework was grounded in the Technology Acceptance Model (TAM) and Theory of Planned Behavior (TPB). Logistic Regression outperformed Random Forest, with an AUC of 0.700 and an F1-score of 0.682. Frequency (β = 0.16, p < 0.01), Sentiment (β = 0.87, p < 0.001), and Engagement Index (β = 0.22, p < 0.01) were significant predictors of conversion. Channel type moderated the relationship between sentiment and conversion, with stronger effects observed among app users. Random Forest highlighted Recency, Sentiment, and Monetary Value as key features. Integrating structured and unstructured data enhances predictive accuracy and reveals nuanced drivers of customer behavior. The moderating role of channel type underscores the importance of context-specific engagement strategies. The framework provides actionable insights for retailers to optimize personalization, allocate marketing efforts by channel, and design predictive systems that adapt to customer behavior dynamics.