Use of K-Means Algorithm in Model Improvement Production Data Grouping for Determination Convection Production Strategy

This research was conducted to implement the K-Means Clustering algorithm in grouping convection production data to support the improvement of efficiency and effectiveness of production strategies. The data used is PT ABC's production data, which consists of important attributes, such as Production ID, Production Date, Product Name, Product Type, Color, Size, Raw Materials, and Order. The research method follows the stages of Knowledge Discovery in Database (KDD), which includes selection, preprocessing, transformation, data mining, and evaluation, so as to ensure that the data processed is relevant and ready to be analyzed. The grouping process is carried out using the K-Means algorithm, which groups data based on attribute similarity by determining the optimal number of clusters. The evaluation of the clustering results was carried out using the Silhouette Score and Davies-Bouldin Index metrics, where the results showed values that represented good cluster quality. A high Silhouette Score indicates that the data in the cluster has good uniformity, while a low Davies-Bouldin Index indicates a clear distance between clusters. The results of the grouping produce three main clusters that illustrate different production patterns, such as clusters with high, medium, and low order quantities. This analysis of the cluster provides important insights in supporting strategic decision-making, such as prioritizing resource management in high-order clusters and evaluating production efficiency in low-order clusters. This research is in line with previous literature that shows that the K-Means algorithm can be used effectively in big data grouping to support strategic planning. The practical contribution of this research is to help convection companies in understanding production patterns, so that production strategies can be designed more efficiently, responsively, and directionally. For further research, it is recommended to add new variables, such as production costs or work duration, as well as test other clustering algorithms to obtain more comprehensive results.