The usage of social media is becoming more and more widespread among business owners to acquire business value. To extract valuable information from the collected data, it is recommended to use business intelligence tools. In my paper, I analyse this business problem both on a theoretical and a practical level. My main goal is to get a result from the data mining techniques utilized, which can be used in order to give useful business advices to the certain business owners.
For my work, I used a database from Yelp, a location-based social media site, which helped me build and analyse two of the data mining models introduced in the theoretical part. One of them is a text mining model that processes the reviews of businesses, and then classifies them into different categories, in two ways. Here, I built an emotion and a polarity classification model based on sentiment analysis. The other data mining technique that I used is based on spatial analysis. With the help of this model, we can analyse the data of the neighbouring businesses. Moreover, based on these two models – sentiment analysis and spatial mining – I created an algorithm that gives a quality score to the businesses based on social network data.
The other part of my thesis work included building a classification and a regression analysis based on the results of the previously created models. To do this, I used and compared 4 classification and 4 regression methods. Then at the evaluation of the models, it can be seen that all of the regression models resulted in better performance metrics than the classification analysis.
Furthermore, at the end of my paper, I introduce how businesses can use the information extracted by the data mining techniques. For this, I created the plan of a user interface that can help business owners to get an overall picture of their and the neighbouring businesses. There are many cases when a business owner would like to know how good or how bad his business is compared to other nearby venues.