International Journal on Recent and Innovation Trends in Computing and Communication ISSN: 2321-8169 Volume: 5 Issue: 7 777 – 781 _______________________________________________________________________________________________ 777 IJRITCC | July 2017, Available @ http://www.ijritcc.org _______________________________________________________________________________________ A Hybrid Approach for Recommendation System based on Web Mining Gurleen Kaur Amanjot Kaur CSE & IT Deptt. Assistant Professor, CSE & IT Deptt. BBSBEC BBSBEC FatehgarhSahib,Punjab,India Fatehgarh Sahib, Punjab, India gurleenkour1122@gmail.com amanjot.kaur@bbsbec.ac.in Abstract - The significant issue of many on-line sites is the introduction of numerous decisions to the customer at once; this for the most part brings about strenuous and tedious in finding the correct item or data on the site. In the traditional methodologies, KNN based classification strategies were utilized which depended on suggestion handle. These have some real issue if the information differs. The arrangement approaches that were utilized as a part of customary work are fit just if the data variation is inside the cluster that they have. However, in the event that the data goes out of bound it is hard to perform classification. In this way, there is a need to include a classifier approach that can work in such conditions. For this, a hybrid approach comprising of Multi-Layer ANN and k-NN is proposed in order to take proper choices if there should be an occurrence of data variation. The proposed idea introduces an intelligent approach which captures the clients going out of bound and adds them into the cluster, so that they can be recommended to the user and no client is skipped. Keywords-Web Mining, Multi-Layer Artificial Neural networks (ANN), K-Nearest Neighbor(k-NN) __________________________________________________*****_________________________________________________ I. INTRODUCTION Data mining methods give users a new energy to look into and control the current vast volume of information. Data mining process finds fascinating data from the concealed information which can either be used for future forecast or potentially insightful outlining the subtle elements of the information [1,10]. Web mining innovation is developing field of data mining for WWW based data and assets. The fundamental center of web mining is to use data mining ways and calculations to remove valuable and concealed patterns from unstructured and tremendous web information or assets [2,11]. Recommendation system uses the selected items as response and recommends the user a number of items that have chosen other users. This problem can have two interpretations: 1) the problem of item‟s recommendation for the user who is already working with the advisory system. 2) The problem of recommendations of objects for the new user that implements the first login [3,12,13]. In this work second problem is there when system can watch clients/users route conducted by following up on the client's click-stream information on a RSS reader site, to prescribe a proper arrangement of items that fulfills the need of a dynamic client in a Real-Time, online premise. The RSS (Really Simple Syndication) reader site is an example of online recommendation system where users can able to read daily news online across the globe. In many cases K-Nearest Neighbor classification technique was used as it is extremely efficient and dependable strategy to know client‟s conduct, behavior and interest at a specific session. But in this paper the hybrid of K-Nearest Neighbor (k-NN) with Multi-Layer Artificial Neural Network is done under which two ways are analyzed, got more exact outcomes and that helps in increasing the effectiveness of the framework. This helps to give exact information to the users for a specific information. The MatLab software was used to interpret and present graphical results. II. RELATED WORK Adeniyi et al. [4]presented a study of automatic web usage data mining and recommendation system based on current user behavior through his/her click stream data on the newly developed RSS reader website, in order to provide relevant information to the individual without explicitly asking for it. The k-NN strategy has been prepared to use on-line and in Real-Time to distinguish customers/guests click stream information, coordinated it to a specific client gathering and suggested a customized perusing alternative that address the issue of the particular client at a specific time. To accomplish this, web clients RSS address document was extracted, scrubbed, arranged and gathered into significant session and information data mart was created. The outcomes demonstrated that the k-NN classifier was straightforward, steady, direct and basic as contrasted and different methods. Bellaryet al. [5] discussed various machine learning approaches used in data mining. Further they distinguished between symbolic and sub-symbolic data mining methods. After that, a hybrid method with the combination of Artificial Neural Network (ANN) and Cased Based Reasoning (CBR) in mining of data was proposed. Bloggers are one of the powerful instruments of web which are considered as one of the significant tool of social and intuitive capacities in making IT world awesome. Two strategies were utilized by Farhad et al. [6] i. e. k-NN and ANNs. These strategies are grouped in light of Kohkiloye and Boyer-Ahmad province bloggers dataset. Considering the k-