Categorizing Multi-Label Product Questionnaires through SVM Based Click stream

Abstract

In this paper, Question Categorization (QC) has been studied most primarily in order to understand customers' search intention. In both of these searches, the items in the question list relate to the category label belonging to the taxonomy tree that is being examined. Despite this, search queries about the product usually vary depending on what is vague, and introduce new products over time, seasonal trends and narrow. Traditional supervised approaches to E-Commerce QC are not possible due to the high volume of traffic and high cost for manual annotation in E-Commerce search engines. Here, clickstream data is utilized to determine the effectiveness of a channel's marketplace. So, using the customer's click concept, to collect large-scale question categorization data, this paper uses unsupervised methods that means SVM algorithm is mainly used in this system. Here the data is in the multiclass and multi-label classifier is used to classify them. This paper gets on a large multi-label data set with specific and individual queries from a specific category. In this paper, a comparison of different sophisticated text classifiers is viewed. This paper calculates the micro-F1 scores of top and leaf, which are considered to be a linear SVM-ensemble.