Predicting the Trend of Indonesian Minimum Regional Income Using Statistical and Deep Learning Approaches

Analyzing economic growth indicators is crucial in determining a country’s economy in the future. One key indicator is minimum wage, in which this paper examined further by using both statistical methods and machine learning methods to forecast the indicator’s future values. The data used is from Indonesian Central Bureau of Statistics’ public website. Utilizing and comparing between four models: Linear Regression Model (LR), Autoregressive Integrated Moving Average Model (ARIMA), Deep Learning Model with Long Short-term Memory (LSTM), and Deep Learning Model with Gated Recurrent Unit (GRU), the paper concluded that GRU had the best result upon evaluation through finding all the models’ Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination (R
²
score), in which GRU had values of 122,701.02, 104,442.75, and 0.3777 respectively. Prediction by GRU showed that Indonesian minimum wage would start to decrease and become stagnant in the 2023-2027 duration. This prediction however could potentially be false due to the GRU model having low R
²
score, thus indicating the model had low predictive ability. Further research would be needed to explore larger datasets, utilize more economic indicators, and trying other models.