Abstract

This study addresses chili price volatility, an important concern that impacts the national economy and societal welfare. Fluctuations in chili prices in the retail market greatly influence market demand, thereby influencing farming decisions, especially chili cultivation. To help make better decisions, Researchers use forecasting, which is defined as the projection of future trends based on the analysis of historical data, using statistical methods. The K-Nearest Neighbor (K-NN) algorithm is used because of its resistance to high noise on large training datasets. However, challenges arise in determining the optimal value of 'k' and selecting related attributes. To overcome this, Feature Selection is applied to refine the model by removing irrelevant features, resulting in a significant reduction in the model error rate. This improvement indicates an increase in the efficiency of the K-NN algorithm with the incorporation of Feature Selection. Our findings show that the model, with backward elimination in Feature Selection, achieves a Root Mean Square Error (RMSE) of 0.202, outperforming the model using forward selection. The prediction accuracy of this model reaches an average of 78.86%, which is much higher than the baseline data of 50%. This shows the success of the proposed method in predicting chili prices.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call