Prediction of Cell Phone Sales from Online Reviews Using Text Mining

Abstract

In a world that is constantly shifting more and more towards a digital realm, having data available for most fields is becoming all the easier. Since all online interactions are automatically recorded, they can be readily accessed to provide valuable data. This helps in tracing and predicting consumer behavior which can help predict the commercial success of various day to day products. One way to obtain usable data from online sources is by using text mining which is a data science technique which analyses textual data and sifts through it to obtain valuable chunks which can be used as an insight without having to convert it to numerical format. Text mining can be done using various tools such as Python and R and further operations can be performed on the data to produce more valuable insights. This paper describes a technique of text analytics on mobile online reviews to discover significant text patterns that exist in the documents. These documents are considered as unstructured textual data. The paper proposes a framework that consists of 3 stages (i) data collection (ii) document preprocessing (iii) text analytics and visualization and (iv) prediction by comparison. The technique is developed using R text mining package for text analytics experiments. We use the patterns obtained from this textual data to try and predict the commercial success of a new cell phone model by using its online review and calculating similarity with reviews of commercially successful cell phones.