LOCATION-BASED SERVICE IMPLEMENTATION TO MATCH CUSTOMER NEEDS BASED ON PREVIOUS BEHAVIOR

ATHEYA, MICHAEL BRANDON, Sigit Priyanta

2016 | Skripsi | FMIPA

The focus of the study is to classify the characteristics of category’s buying behavior using CHAID algorithm and implement the result using location-based service. To do so, several consecutive phases were conducted; they are pre-processing, labeling, classification, Analysis, and implementation. The dataset originally contained record more than 24,000 customers and more than 32,000 transactions around 1st August, 2015 – 30th October 2015 in Grand Indonesia, Jakarta. In preparing the data, the model used 6 features such as age group, gender, credit card group, amount of transaction group, buying frequency in the previous 2-3 months and buying behavior in the last 1 month. Then, CHAID algorithm was applied to the datasets. As the result whitelists are made and ready to be implemented through the LBA core. Experiment shows that the accuracy of the classification depends on many factors such as number of features, total dataset, transactional log period, and varieties of transaction. This conclusion can be seen though validation process where groceries category has the biggest accuracy (94.64%), followed by Food and Beverage category (75.759%) and fashion (73.454%). The model created in this research worked in both telecommunication and bank system.

Kata Kunci : data mining, classification, CHAID Decision Tree Algorithm, Location-Based Service, Transaction Historical Dataset.