Estimación de la velocidad de producción de una máquina mezcladora mediante la implementación de modelos de Machine Learning

Abstract

This project addresses the challenge of predicting the speed of a mixing machine in an animal feed processing plant. The primary objective of this work is to implement machine learning models to estimate the mixing machine’s speed. Initially, data preprocessing was conducted, including cleaning and filtering outliers to enhance data quality. Subsequently, SARIMA (Seasonal Autoregressive Integrated Moving Average) modeling was applied for time series forecasting. Additionally, Long Short-Term Memory (LSTM) neural network models were adjusted to predict the speed for each product. The project results showed that, while the SARIMA model was suitable due to seasonality in the data, the LSTM networks demonstrated superior performance in predicting mixer speed, reflected by approximately 84% of products showing lower MAPE values. Consequently, the LSTM model was selected and used for implementing the MLOps pipeline. Additionally, exploring a mixed modeling strategy that combines SARIMA and LSTM approaches was recommended to maximize accuracy. This approach can optimize efficiency and enhance the quality of production activities at the plant. Keywords: ARIMA Models, LSTM, Data Management, Time Series