Temperature Forecasts in New York City Exploiting ARIMA Model

Efforts to estimate air temperatures play an important role in predicting climate impact on ecological, industrial, and agricultural fields. Accurate temperature prediction benefits the government, industry, and the public for planning activities. Moreover, the temperature forecasts provide people with safeguard life and property. This paper conducts an exploratory data analysis of the temperature time series data, taking New York City as a representative. The time series data contains the daily temperature in New York City from 1995 to May 2020. The considerable time scope is so comprehensive that it makes a forecast based on the periodic performance similar in approximately 20 years. In this work, the autoregressive integrated moving average (ARIMA) model is leveraged to forecast future values based on past periods. Its hyperparameters are determined by autoregressive correlation function, partial autoregressive correlation function, and Bayesian Information Criterion. The ARIMA model assumes that the future performance will resemble the past, and the temperature in different time stamps is similar, which makes it a suitable model in this case.