Temperature mining using spatio-temporal based neuro system

Data mining is the search of effective patterns that exist in a huge dataset. In this paper, we introduce four novel features, entropy, joint entropy, cumulative frequent itemset and weighted feedback that are used as input of neural network. Two types of neural network, for example, back propagation and probabilistic neural network is used for predicting the average, minimum and maximum of monthly temperature of Dhaka, Bangladesh. The accuracy of prediction by different neural network architecture is presented by varying different parameters of neural network. Our goal is to find out the neural network that can optimally perform the prediction task. Results demonstrate that probabilistic neural network performs better prediction compared to back propagation neural network except for few cases.