International Journal of Computer Applications (0975 – 8887) Volume 120 – No.5, June 2015 13 A Neuro-Fuzzy Integrated Clustering for Weather Knowledge Analysis Sonakshi Dahiya CSE & IT Dept. ITM University Gurgaon, India Yogita Gigras Astt. Professor,CSE & IT Dept. ITM University Gurgaon, India ABSTRACT Weather Information processing and knowledge extraction is one of the challenging applications of data mining. This process area requires authenticated and intelligent processing to obtain accurate information from the knowledge set. In this work, an intelligent clustering mechanism is defined to acquire such information. This neuro-fuzzy based model is applied on raw dataset defined with various weather characteristics including humidity, temperature, rainfall etc. The work is divided in three main stages. In first stage, the filtration over the dataset is performed to get more relevant information set. In second stage, the clustering is performed to divide the information set in knowledge groups. In final stage, the filtration over the knowledge set is performed to acquire the most effective knowledge. The results show the effective information analysis is obtained from the work. Keywords Neuro-Fuzzy; Weather Forecasting; Knowledgeset; Clustering. 1. INTRODUCTION Clustering is one of most common data processing activity defined to generate the data patterns. These data patterns are considered as the measurement vectors defined in multidimensional space so that the relevant information can be collected in different data groups. The formation of these clusters is based on the similarity analysis. The clustering is effective to provide the data filtration as well as data reduction so that the information discovery can be done in an effective way. The cluster formation is based on the association analysis and structural analysis. This group generation and formation is based on the evidence analysis applied on the constraint specification. In this present work, a process variable based analysis approach is defined to divide the dataset in various data groups. The work is here defined based on multiple parameter or objects. In this work, the weather forecasting dataset is taken for more than 20 years. The information set considered here is defined with various associated features including the humidity, temperature and rainfall. All the dataset values are taken at different time instances obtained throughout the year. This dataset is collected for a specific city. In this work, the cluster analysis approach is defined to analyze the behavior relation between these data values. This relationship establishment is also based on multiple vectors including the distance, variation, similarity measures. In this work, clustering effective classification model is defined to divide the dataset in certain groups. The pattern analysis is the key feature considered in this work for relevant class formation. This kind of information processing is the specialized data mining application defined under time domain. To acquire the information from such dataset, the time series segmentation is required. 1.1 Time Series Segmentation This kind of segmentation process is considered as most intelligent and relevant information processing system defined for real environment. This kind of information process uses the neuron specific model to train the information set and adapt the required information from it. This segmentation process is defined in 2D space with specification of neurons to form the clusters. The learning model is here applied to generate the data patterns so that the neighbor information and distribution processing is performed. Such kind of data processing includes the cluster formation on larger dataset. The topological ordering is performed to identify the similarity between the data groups and this similarity analysis is handled by the neuron specific constraints. The intelligent neuron processing is defined to identify the cluster requirement at early stage and later on the data segmentation is applied. As the process continue, the cluster splition and merge operations can also be applied in an integrated way. This kind of segmentation is also defined under specification of clustering model so that that the data division and the reduction is performed in effective way. The data patterns are generated based on the temporal specification so that the dataset formation and the relative information switching can be obtained from the work. This kind of information processing also defined the analysis as the integrated stage to the process. This processing model includes the analysis applied at the earlier stage so that the information division can be performed on larger dataset. In this work, the year specific time series segmentation is applied. The optimization is here been done using neuro fuzzy based rule specification. 1.2 Neuro-Fuzzy In this work, neuro-fuzzy model is been applied to acquire the most effective information from the dataset. The layered model is here defined to process the time series data and to obtain the effective information results. The model used the filtered dataset as the input neuron set for the process. The constraints are here defined under the specification of forecasting features. The complex information format is defined with informational view to perform data modeling. Based on this, the data mapping is performed under constraint specification. Along with this, the predictive decision vectors are defined along with the specification of training time, validation rule etc. Based on this learning rule specification, the effective information validation is done. Finally the learning process is applied under fuzzy rules to acquire the information and to present the results.