Web-based fuzzy time series for environmental temperature and relative humidity prediction

In this research, we develop a web-based instrument for air temperature amplitude (T) and relative humidity (RH) predictions using internet connection. There are decisive factors of climate on Earth related to climate change and global warming. Additionally, air temperature amplitude and relative humidity play important roles for the health of the Earth. We developed a wireless sensor system that transmits data to the web server via online internet connection. Semiconductor air temperature and humidity sensors are installed in remote terminal stations. Acquired data were transformed from analog to digital using a converter circuit, and the results were stored in a local database. A computer application is developed to send data online and in real time to a web server using an internet modem. Transmitted data are received and saved on a web hosting database service. For prediction of air temperature amplitude and relative humidity, a Fuzzy Time Series Algorithm induced by Markov transition matrix is used and is pre-installed in the web server. Performance testing for air temperature amplitude and relative humidity measurements showed encouraging results that correspond well with correlation coefficient (R) close to 1. Those testing revealed values R of temperature is 0.9987 and R of RH =0.9946, with minor errors of T=0.18 Celsius degree and RH=1.43%. Results of temperature amplitude and humidity predictions are displayed real time on the dashboard for one day ahead. Testing results also indicated major deviations between actual and predicted data whenever there are sudden fluctuations in the readings of temperature amplitude and relative humidity. Subsequent test results showed that the Fuzzy Time Series-Markov model has an error of 4.6% for temperature and 2.76% for relative humidity prediction, respectively. When further calculations using mean absolute percentage error (MAPE) were carried out, these error values further varied, depending on the number of data and the data characteristics themselves.