Previsão de séries temporais de sensores de temperatura da empresa Nexxto

Abstract

The main objective of this work is to develop a time series forecasting model based on the temperature sensors data of the company Nexxto to provide a simulation for the company’s product development team. The Nexxto’s platform offers temperature sensor data monitoring, as well as reporting and alarming functions for healthcare customers. This work was developed using data analysis tools in Python programming language to perform processing and statistical analysis of sensor data. The treatment process prepared the data for the autoregressive integrated moving average (ARIMA) model implementation to predict sensor values. Analyzes were performed to define the forecasting model parameters based on the evaluation metrics. The company Nexxto already uses a sensor data simulation tool. However, using the model developed in this work allows simulating the real behavior of the sensors installed in customer environments. The model results showed errors below half a tenth of a degree Celsius and an acceptable runtime for the application.