Vtt Information Technology

This report examines the problems of customer relationship management (CRM) particularly customer segmentation and customer profiling, and how data mining tools are used to support the decision making. We first describe the steps towards predicting customer’s behavior, such as collecting and preparing data, segmentation and profile modeling. Then, we present a general overview of most used data mining methods including cluster discovery, decision trees, neural networks, association rule and sequential pattern discovery. The report also covers a discussion about Web mining which is treated as a separate section due to its current popularity in electronic commerce. A guideline to choose a data mining software package is also given in the last section. LOUHI DATA MINING FOR CUSTOMER PROFILING 1 VTT Information Technology Last modified on 02.08.01 ii Preface This report is written within the frame of LOUHI project. The purpose of this report is to provide to our project partners a general view on data mining methodologies used in today’s customer relationship management (CRM). In today’s marketing strategies, customers have a real value to the company. Therefore, it is essential to any company to be successful in acquiring new customers and retain those that have high value. For this, many companies have gathered significant numbers of large and heterogeneous databases and these data need to be analyzed and applied in order to develop new business strategies and opportunities. The problem is that "we are rich in data and poor in information". What are the methods that can be used to automatically extract knowledge from data? Recently, new data analysis tools have appeared using various "machine learning" techniques. It is through the use of machine learning that data mining tools emerge. The advantage of data mining is that it can handle large amount of data and "learn" inherent structures and patterns in data; it can also generate rules and models that are useful in replicating or generalizing decisions that can be applied to the future cases. Data mining tools are therefore very useful in market segmentation, customer profiling, risk analysis, and many other applications. The growing interests in data mining tools have also fostered the growth of the data mining tool market. Nowadays, there are so many vendors offering all range of products. For a person who is new in the field and would like to use those tools, it is essential that he/she understands how each method works in order to be able to choose the adequate ones for his/her problems. This report aims tempts to make the reader familiar with the most used data mining methods (clustering, decision trees, neural networks, associations...) and also the emerging Web mining techniques. We do not intend to provide a complete overview of each technique, neither to compare systems. LOUHI DATA MINING FOR CUSTOMER PROFILING 1 VTT Information Technology Last modified on 02.08.01 iii