Prediction is a phenomenon of knowing what may happen to a system in the next coming time periods. Weather is a time series based, continuous, data-intensive, dynamic, and chaotic process.Due to dependence of weather on time series based data and non-linearity in climatic physics neural networks are suitable to predict meteorological processes. In the present research, firstly weather related data have been collected, weather parameters have been selected, N-Sliding window technique is applied, relations between dependent parameters are found and data has been normalized to feed to the network as input. After the per-processing of data, suitable neural network architecture has been determined and then the network has been trained by feeding the input as well as output data set under supervised training. Afterwards, testing of the networks has been done for different input sets to check how accurately the network has been trained. Finally, a comparison between the existing and proposed time series based technique has been done. The proposed hybrid technique can learn efficiently by combining the strengths of genetic algorithm with back propagation algorithm.