A comparison between the simple linear regression model and time series models in forecasting the number of cancer cases at the National Institute for Cancer Treatment in Misrata during the period from 2012 to 2017.

Abstract

This study aims to compare the simple linear regression approach and time series methods and to select the most appropriate model based on statistical tests and various selection criteria for forecasting the future number of cancer cases. The data showed a strong positive correlation of 0.88 between time and the number of patients, indicating that the data could be represented by a straight line. Accordingly, a simple linear regression equation was developed, and the model demonstrated a relatively good fit, with a coefficient of determination (R²) of 78%.

The cancer incidence time series was found to be non-stationary and exhibited a general trend. The most suitable model for this series was identified as the first-order integrated autoregressive moving average model ARIMA(1,1,1). A comparison was conducted between this model and the simple linear regression model for forecasting the future number of cancer cases. The results indicated that the ARIMA(1,1,1) model outperformed the simple linear regression model based on statistical tests and other evaluation criteria.

This conclusion was further supported by the in-sample predicted values from the ARIMA(1,1,1) model, which showed a high degree of closeness to the actual values, unlike the simple linear regression model, whose in-sample predictions followed a straight-line pattern and were therefore less consistent with the actual observed values.