ارايه‌ي روشي نوين در خوشه‌بندي بازار خدمات تلفن همراه، با استفاده از داده‌كاوي (مطالعه‌ي موردي: بازار موبايل)

A Novel Approach to Market Clustering using Data Mining: Case Study: Iranian Mobile Market

هدف اين مطالعه شناسايي بخش‌هاي متمايزي از مشتريان بازار موبايل كشور با كمك روش‌هاي كمي است. امروزه با توجه به اهميت داده‌ها در ارايه‌ي مناسب‌تر خدمات، بازاريابي هدف‌‎‌مند و تخصيص بهينه‌ي منابع، «داده‌كاوي» به‌عنوان يكي از ابزارهاي شناسايي الگوهاي موجود بين حجم عظيمي از اطلاعات درهم ريخته، اهميت يافته است. از آنجا كه در صنعت تلفن همراه، دسترسي به حجم بالايي از اطلاعات مفيد كاربران به‌آساني ممكن است، مي‌توان از اين ابزار استفاده ‌كرد. يكي از كاربردهاي اين ابزار «خوشه‌بندي مشتريان» است. در اين مطالعه ضمن شناسايي و گردآوري، خوشه‌بندي و تحليل داده‌ها، با كمك نرم-افزارSPSS Clementine و الگوريتم دومرحله‌يي سري زماني، داده‌هاي قبوض مشتركين تحليل شد. نتايج خوشه‌بندي حاكي از وجود پنج خوشه‌ي متمايز است. با بررسي درآمد ناشي از گروه‌هاي پنج گانه ، برقراري اصل پارتو مشهود است. همچنين مشتركين در خوشه‌هايي كه ظاهراً سهم كمي از كل مشتركين را تشكيل مي دهند، تا 20 برابر مشتركينِ بزرگ‌ترين خوشه‌ي شناسايي شده، درآمدزايي دارند.

The purpose of this study is to determine clusters of mobile subscribers, such that members within a cluster demonstrate common characteristics, while, at the same time, distinctive features can be uncovered among members of different clusters. Along with the increasing importance of data for organizations, it seems critical for managers to extract valuable information out of unstructured data through efficient access, sharing, extracting and implementing the data. Data mining is a promising tool to dig out patterns from a large amount of unsorted data. Telecommunication industry benefits from such data, opportunity. One of these favorable applications is consumer clustering, which leads to market segments. The study was conducted in three successive phases. First, the required data was determined and gathered through multiple expert panels. The second phase included running the clustering process, during which, the records were clustered. The last , but not the least, is analyzing the results. For the first phase, 16 million mobile post-paid bills were analyzed, using a two-step algorithm in SPSS Clementine. Each bill included 11 variables, out of which, 7 variables were chosen as important and effective for clustering and marketing purposes, due to technical and managerial insights. The other 4 variables were filtered. The clustering process was applied using these 7 variables as the input to the two-step clustering algorithm. The results revealed 5 distinctive clusters that demonstrated different behavior in various aspects, including the Pennywise, who just use limited voice call service, the Traditionalias who use a mid-level of voice service and a little of SMS, the Exorbitants use the maximum level of voice and medium level of SMS, pioneers who use newly offered services and provide the maximum revenue per user, and the followers, who are the biggest SMS users and use new services after pioneers. The Pareto principle was obviously observed when scrutinizing the earned revenue from each cluster. Dividing the revenues from each cluster by its size, we were able to plainly note that each member of the smallest cluster generated revenue 20 times greater than that of the largest one.