Hareketli en küçük kareler modeli önerici sistemler: Kredi kartı pazarı üzerinde bir uygulama

Following the introduction of MapReduce and Apache Hadoop,it has been possible to process immense datasets that are beyond the capabilities of traditional database management system techniques. This created a new area of study: Big Data. Big Data is generally used to define the massive and unstructured datasets, unsuitable to process with subject traditional methods. Growing interest on data intensified on the areas where it is available the most: e-commerce businesses, movie review sites, music player platforms to name a few, where user interaction is digital so that it can be logged and traced. Practices mainly aim analyzing user profiles, predicting preferences and making appropriate recommendations. Though it is relatively easier to analyze feedbacks and predict preferences in these cases, where user ratings, scores, favorites or likes/dislikes are available, the bigger part of the value lies within the indirect data, as direct feedbacks are usually not in grasp. Businesses should harness any information available and build proper correlations to feed the recommendation system. In this study, credit card transaction logs will be studied to predict card holder's next transaction sector and propose marketing offers correspondingly. I hope it will shed light on future researches on recommendation systems with implicit data.