Analyzing visitor trends to optimize data-driven strategies in Pusat Sains & Kreativiti Terengganu

This project aims to implement predictive analytics to assist Pusat Sains & Kreativiti Terengganu (PSKT) in addressing three key challenges which are limited visibility into visitor trends and demographics, underutilization of data in decision-making, and the inability to identify peak visitor periods. To overcome these issues, historical visitor data from January 2022 to December 2024 was collected and analyzed using the CRISP-DM methodology, which guided the project through business understanding, data preparation, modelling, and deployment. Three predictive algorithms including Decision Tree (DT), Random Forest (RF), and Naive Bayes (NB) were tested to classify visitor levels into low, medium, and high categories. Among these, the Random Forest algorithm with a 70:30 data split achieved the highest accuracy of 78.48%, making it the most suitable model for integration. The predictive results were visualized through a Power BI dashboard consisting of five main pages, Home, Total Visitor Overview, Visitor Demographic, Visitor Prediction, and Visitor Impact. These visualizations enable PSKT management to explore visitor trends, understanding visitor demographic and assess the influence of external factors such as events and rentals. Expert evaluation confirmed the dashboard's clarity, usefulness, and potential as a decision-support tool. However, limitations were identified, including imbalanced data, wide value ranges, and the dashboard's lack of real-time data integration. Recommendations for future enhancement include expanding the dataset to over five years and enabling live data updates to improve prediction accuracy and extend dashboard lifespan. Overall, this project contributes a practical and scalable solution not only for PSKT but also for other institutions seeking to leverage predictive analytics for improved visitor or customer trend analysis.