A Comparison of Prediction Algorithms in Food Sales with Different K-Folds Cross-Validation

Forecasting in business is important because it enables businesses to deliver the ideal quantity of items at the ideal moment. According to cross-data studies, there is a significant link between market forecasting and business growth. Therefore, intelligence system such as prediction is crucial. Many algorithms can be used for prediction. It is important to find out the optimum approach with the lowest error. This paper compared three main algorithms, namely Support Vector Machine (SVM), Linear Regression (LR), and the Autoregressive Integrated Moving Average (ARIMA). The experiment was carried out using food sales datasets over 3 years. The experiment was carried out using food sales datasets over 3 years. Each algorithm was evaluated for its error values with RMSE, MSE, and MAE. The results showed that the SVM algorithm is better than the ARIMA and LR algorithms in prediction. The SVM algorithm was able to predict sales numbers with an accuracy of 83% and has an RMSE error of about 17%. In addition, in building prediction models, especially for food data, based on an error of each k-fold cross-validation, the use of k > 5 worked better in food sales forecasting for all algorithms used.