層級貝氏模型應用於來電答鈴廣告之會員價值之研究

Abstract

本研究試圖分析答鈴廣告之會員在播放不同類型的廣告時，所產生之不同個廣告效益，將每個會員皆視為獨立之異質性個體，透過分析過去廣告播放期間所產生之通話記錄及廣告訊息之數據資料，建構出一個會員價值之衡量模型。研究主旨在於探究會員之通話結構對於不同類型廣告之價值表現。並且利用層級貝氏統計分析方法以分析會員之個人化變數對於會員價值的關聯性。 本研究運用了國內某家來電答鈴廣告商之通聯及廣告資料庫，利用其需要按特定按鍵以獲取進一步資訊之廣告特性，以特定按鍵產生之數量來衡量會員價值。變數除了既定的欄位變數外，另外加入了社群集中度、週末社交率等社會結構變數，將資料分成兩層，第一層之自變數隨著不同廣告而改變，第二層變數只隨著不同會員而改變，並運用層級貝氏統計分析及集群分析來探究以下兩個主題： (1) 建構一個會員價值衡量模式，研究哪些變數對於哪種廣告會有哪些不同的影響。 (2) 利用某些變數及特徵將會員分成若干群體，研究需要具備哪些變數特性的會員來播放廣告，以期達到更好的廣告效果。 本研究可以提供企業在廣告排程及選擇會員上能有更具效率及效果，能夠在招募會員及優化排程上作出貢獻。

Our research is aimed at constructing the valuable module of membership by analysis the record of phone communication and data base generated from advertisement as viewing each membership as the independent heterogeneity. It believes that playing the different type of advertisement will impact on the profit, which means that the valuable advertisement will bring in more profit. Therefore, the goal of research is diving the constructer of phone communication of membership versus different type of AD. We also find the correlation between the variation of membership and value by hierarchical Bayesian analysis. Cooperating with the telecom server who the database of advertisement and the feature of getting information by certain key, we can get the useful information by pressing on certain key. By counting it, we can evaluate the value bringing by each member. The variable is including not only certain column variable but also social structural variable such as the centralization of group, the Weekend social rate etc. the data is divided into two layer, one is the variable is changed from ad, another is the variable changed from member. We figure our the below topic by using the hierarchical Bayesian analysis and cluster analysis. (1) Constructing the mode for evaluating the value of membership and studying the impact on advertisement when changing the certain variable (2) dividing the member into several group by the certain variable and the characteristic of membership, we are aimed at to find our the certain characteristic, the membership who equipped with those characteristic can bring in more profit when playing the advertisement. Our research can provide the cooperation with the better efficiency as they schedule the ad and choose their valuable membership. Based on my research, it is believed that it can make the contribution on optimization of scheduling ad and recruiting membership.