<![CDATA[The use of online marketplaces is rapidly expanding in Indonesia, particularly within the fashion industry. To develop effective marketing strategies, it is essential to understand consumer behaviour through customer segmentation. With a deeper understanding of consumer behaviour, XYZ company, which is engaged in the fashion industry, can improve the effectiveness of marketing strategies and respond to consumer needs more accurately to achieve a significant increase in sales. This study aims to implement a customer segmentation model using clustering methods with machine learning algorithms, specifically K-Means and DBSCAN, following the CRISP-DM Data Mining Framework for data processing. The research utilizes purchasing transaction data from XYZ fashion industry, applying pre-processing techniques such as Standard Scaler and PCA before clustering. The K-Means and DBSCAN algorithms are implemented and evaluated using Silhouette Score and Davies-Bouldin Index matrices. Results show that the K-Means algorithm outperformed DBSCAN, achieving an optimal cluster number of k=7 with a Silhouette Score of 0.549 and a Davies-Bouldin Index of 0.593, compared to DBSCAN's Silhouette Score of 0.29 and Davies-Bouldin Index of 0.92. The final implementation involves creating a dashboard that automatically processes data and generates clusters to support customer segmentation decisions. The model was deployed through a simple website using FastAPI for backend Python execution and React with TypeScript for the front end. Future studies could address limitations by incorporating recent datasets to improve model accuracy, exploring alternative algorithms like Gaussian Mixture Models (GMM) for additional insights, and focusing on robust deployment strategies for real-world applications within the fashion industry.]]>
Copyrights © 2025
