Improving Customer Value Index and Consumption Forecasts Using a Weighted RFM Model and Machine Learning Algorithms

Introduction

The game industry spans ages and countries and is expanding through game-related platforms such as e-sports competitions and live games. In the context of the outstanding performance of key game products of head game companies; the gradual process of new version approval release; the rapid development of cloud games after the commercialization of 5G as well as the development of VR/AR games - quickly gaining a player’s favor with an attractive marketing strategy is very important in this highly competitive industry.

Gartner (2014) defined big data as a massive, high-growth, and diverse information asset that requires new processing models to have greater decision making, insight and process optimization capabilities. The famous futurist Toffler (1980) portrayed “big data” as “the third wave of cadenza” in his book “The Third Wave”. Today’s big data development is expanding globally with apparently unlimited business opportunities. Having recognized the influence of big data on various industries the major developed economies, such as the United States and European Union, are actively promoting big data strategies. China also regards it as an important support for their new economy. Data mining generally refers to processes of searching for more specific information, hidden in large amounts of general data. Mining is usually based on a database, composed of large amounts of accumulated data, and the process draws out potentially valuable information to support or improve decision making.

Data mining is a powerful technique to help companies identify patterns and trends in their customers’ data, and then drive improved customer relationships; accordingly, it is a well-known tool in customer relationship management (CRM). Based on traditional methods of data mining, Cheng and Chen (2008) proposed a new procedure, joining quantitative value of RFM attributes and K-means algorithms into rough set theory (RS theory), to extract meaning rules. This development can effectively reduce drawbacks in previous data mining tools. At the same time, data mining has a tremendous advantage for researchers, because it enables them to extract further hidden knowledge that has been inherited in the raw data. The study of Ravasan and Mansouri (2015) explained a brand new and practical fuzzy analytic network process (FANP), based on a weighted RFM (Recency, Frequency, Monetary value) model for application in a K-means algorithm for auto insurance customers’ segmentation. Previously, Hsieh (2004) had proposed an integrated data mining and behavioral scoring model, to manage existing credit card customers more effectively in a bank.

As a new field application, game data mining is emerging, with customer segmentation and customer feature mining the most important entry points. Achieving more precise customer segmentation and more effective customer feature extraction is a long-standing technical difficulty in the game industry (Shahri et al., 2019). Clarifying the customer value measurement index, constructing a customer value identification path, classifying customer value, as well as mining the core consumer and key consumer group characteristics are all necessary for better operation. This paper takes the stored value data of a certain game of a famous Chinese game company as the research sample, to undertake the following investigations: use the improved RFM model to score each player’s value index, and classify the players according to the indicators in this model; utilize logistic regression, decision tree, SVM and other models to predict the total amount of stored value of each new player in the future, and evaluate the accuracy of different models; select the best model and optimize it to predict the amount of stored value of players; finally, according to the players’ distribution, heterogeneity and stored value behavior, to provide more reliable reference indicators for corporate marketing and customer relationship management. These studies are designed to ensure that companies and marketers can better understand their customers, and then choose differentiated marketing tools to assist with the business development. From the perspective of market space and policy trends, the overall solution of game operators’ customer segmentation and prediction of future benefits from new customers, can be used as a benchmark that can positively impact the game industry and act as an indicator for other industries.