When there is an excess of carbon, sulfur and any dangerous gaseous emission, then the air is polluted. Breathing through polluted air can cause harm and may lead to several diseases such as respiratory failure and cancer. This study is about the prediction of real SO2 time series through the chaotic approach in Kota Kinabalu, Sabah. Before the prediction model can be built, the existence of chaotic behavior needs to be confirmed. Chaotic behavior can be detected through the phase space plot and Cao method. The observed time series is confirmed to be chaotic. Prediction through improved local linear approximation method (ILLAM) is done. Three parameters are determined which are delay time