The ease of investing in the digital era has driven Generation Z to dominate stock market participation, particularly in blue-chip stocks such as PT Bank Rakyat Indonesia Tbk (BBRI). However, stock price fluctuations influenced by macroeconomic factors, regulations, and global market sentiment make it difficult for investors to make accurate decisions. Decisions based on insufficient information pose a significant risk of loss, especially for novice investors. This study proposes a hybrid LSTM-XGBoost approach for predicting BBRI stock prices, combining the strengths of LSTM in capturing nonlinear time series patterns and XGBoost's effectiveness in reducing prediction errors. The model leverages both historical data and feature extraction outputs from the LSTM model. Future stock price values are then predicted by XGBoost using this combined dataset. The Hybrid LSTM XGBoost model outperforms the individual base models in terms of prediction accuracy, achieving an RMSE of 117.89, MAE of 92.45, and MAPE of 2.21%.

Social and economic disparities between regions in Indonesia are still a serious problem, reflected in the high Gini Ratio and disparities in purchasing power. The Consumer Price Index (CPI) is an important indicator in measuring consumption patterns and price pressures between regions. This study aims to cluster 150 districts/cities based on consumption patterns through the CPI using the Agglomerative Hierarchical Clustering (AHC) and Spectral Clustering methods combined with Principal Component Analysis (PCA). The innovation lies in the comparison of the two clustering methods as well as the application of PCA to clarify the data structure. Evaluation with Silhouette Score and Davies-Bouldin Index showed that AHC gave the best results with four representative clusters (Silhouette: 0.76; DBI: 0.32), compared to Spectral Clustering with three clusters (Silhouette: 0.75; DBI: 0.54). Each cluster has different expenditure characteristics, useful for data-driven policy making. These results show that AHC is more effective in capturing interregional variations in consumption.