Time Series Analysis for Tractor Sales using SARIMAX and Deep Learning Models

Polina, Pooja, Ganesan, Swathi ORCID: https://orcid.org/0000-0002-6278-2090, Karunarathne, Lakmali ORCID: https://orcid.org/0009-0000-7720-7817 and Somasiri, Nalinda ORCID: https://orcid.org/0000-0001-6311-2251 (2024) Time Series Analysis for Tractor Sales using SARIMAX and Deep Learning Models. International journal of computer communication and informatics, 6 (1). pp. 27-57.

Preview

Text
231-Article Text-398-1-10-20240723 (3).pdf - Published Version
Available under License Creative Commons Attribution.
| Preview

Official URL: https://doi.org/10.34256/ijcci2413

Abstract

Time series forecasting is known for playing vital role in many industries to make important decisions and strategies. This study concentrates on providing accurate insights that can help manufactures and stakeholders of agriculture machinery industry on future sales of tractors by applying both traditional and deep learning models like SARIMAX which is extension of SARIMA and deep learning models. Research starts by observing history data which include years of tractor sales then preprocess the data to find its quality and stationarity further applying SARIMAX model to find trends and seasons and cycles in the data and this model is evaluated by famous metrics like Root Mean Squared Error (RMSE).Deep learning models like Gated Recurrent Unit (GRU), Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), Bidirectional LSTM, CNN LSTM Encoder Decoder, Convolutional Neural networks (CNN). They can help in enhancing the forecasting accuracy by handling all the non-linear relationships and their dependencies in the timeseries and this study will provide comparative analysis of deep learning models and SARIMAX model. Where SARIMAX outperformed the deep learning models with RMSE score 0.01 and provide forecast of next year’s tractor sales using SARIMAX model from the study and use q-q plot, residual plots and ACF and PACF graphs to make sure forecast was done accurately.