Finding suitable locations for live campaigns using different machine learning techniques

Abstract

For business booming, in recent years the idea of find an ideal location for reaching potential customers has been the focus of past research. Traditional approaches have faced many negative responses, now to make business profitable the new marketing strategy is live campaigns. With the growth of location-based social networks collecting data of user mobility and popularity of places has recently become attainable, but not without analyzing to find the optimal location and time for live campaigns with greater accuracy. In this paper, we study the predictive power of various machine learning and mining features on finding suitable location for live campaigns through the use of a dataset collected from Foursquare in New York. We selected 10 candidate areas where the data was preprocessed according to the feature, a score is computed on the candidate areas to do live campaigns based on the features using most suitable algorithm with the accuracy. The results with Random Forest and Decision Tree are shown at the end of the report. Lastly, our proposed model shows how performance varies when using different features and predicting the suitable locations for live campaigns. We achieve 88.25% accuracy in Decision Tree regression model and an accuracy of 88.48% and 70.04% in Support Vector Machine (SVM) and Random Forest respectively.