REVENUE PREDICTION BASED ON TIME SERIES DATA USING THE CONVOLUTIONAL NEURAL NETWORK (CNN) METHOD AT PISANG KEJU BATAM 80 MSME

Abstract

Pisang Keju Batam 80 MSME faces challenges in managing sales and production due to fluctuations in revenue across different periods. This condition makes it difficult for the business to estimate future revenue and determine optimal production planning and sales strategies. This study aims to develop a revenue prediction model based on time series patterns using the Convolutional Neural Network (CNN) method. The dataset consists of 2,192 historical sales records. The research stages include data preprocessing, dataset splitting, normalization using Min-Max Scaling, CNN implementation, and evaluation using Mean Absolute Percentage Error (MAPE). The dataset was chronologically divided into 60% training data and 40% testing data. The model was trained for 100 epochs using the Adam optimizer with MSE as the loss function. The testing results obtained a MAPE value of 1.6496%, indicating that the CNN model was able to produce predictions close to the actual values and has the potential to support production planning and sales strategies for MSMEs