Using the Hybrid Approach of Logistic Regression and Support Vector Machine Methods for Air Pollution Forecasting