Abstract:
Weather prediction is an anxious work for researchers and has completed a lot of research
interest in the last few years. It has been a demand application of climatology and one of
the specialized challenging problems in the world. Meteorological Department are
working very hard day by day to improve the accuracy. But none of them become
hundred percent successful. So, it is a continuous process and we decided to involve
ourselves in this challenge. In our case, we have investigated about different techniques
in weather prediction. Finally we intend to use data mining technique in our research.
This paper comes up with a current method to increase a fitness-oriented architecture for
the weather notification systems which predict whether using these data mining
techniques. Our main approach to apply different classification algorithm on a bulk of
realistic data and watch any major difference among their output of prediction. More
precisely we always tried to choose the best algorithm. We observe their behavior using
graphical interface. Most of the case we got Support Vector Machine to fulfil our
satisfied output. Then we have studied how the Support Vector Machine works in
classification and what is the internal infrastructure of it. We were trying to improve this
algorithm and proposed a better version.
