Research On Personalized Marketing Strategies Based On Intelligent Optimization Methods

With the astonishing growth of e-commerce, personalized marketing has become a significantdirection for the satisfaction of customer's specific demands, the innovation of business services andthe improvement of core competence. However, e-commerce's characteristics such as the unlimitedaccessibility, the real-time response, and the intense competition, bring up new requirements for thepersonalized marketing. Companies must identify these characteristics, create compatible marketingmodels and design effective solution methodologies. This is a significant topic because it can helpcompanies develop customer-oriented e-commerce models, create new online services, increasecustomer satisfaction and enhance company profitability.This paper focuses on the personalized marketing strategies and employs the intelligentoptimization methods to investigate the product strategy, promotion strategy, and price strategy. Ouraim is to attract potential customers, increase customer satisfaction, and maximize company profits.First, we investigate the personalized preference of customers and help companies designpersonalized product strategy. Second, we integrate the personalized product strategy and onlinepromotion activities to maximize promotion profits by optimizing the promotion discount and therecommended product portfolio. Third, we investigate the personalized price strategy to improvecompany profits by attracting more customers to buy more products. The detailed contents andinnovations of this paper are as follows:(1) This paper investigates recommendation methods for specific information sources. Ourresearch provides guidelines for companies to improve the efficiency of recommendation algorithmand enhance the forecast ability of recommendation rules. The feature selection tactic based on AntColony Optimization (ACO) is proposed to deal with data reduction problems of the massive data.The associative classification method founded on the β-stronger relationship is proposed to integratethe recommendation accuracy and the forecast abilities of the recommendation rules.Our study shows that the proposed ACO-based feature selection method can discover criticalfeatures from the massive marketing data and provide high quality data sources for the personalizedproduct recommendation. According to the requirement of the recommendation practice, the β-stronger model offers the flexibility for decision makers to set the parameters of our model. Theproposed recommendation model for specific information sources satisfies the requirement ofdiscovering customer preference and is the foundation to develop personalized marketing strategies.(2) This paper investigates the combination and optimization methods of the recommendationdecisions from the multiple information sources. Our research provides guidelines to improvecustomer's satisfaction by utilizing the recommendation results from various information sources.The transformation method from the inconsistent rules to the recommendation evidence bodies isproposed to deal with the conflicting problem of the recommendation decisions. An utility analysismethod is proposed to explore the utility paradox problem when employing Evidence Theory tocombine recommendation evidence bodies. The method of integrating online reviews is also studiedto optimize recommendation results and suggest products to satisfy customer.Our study shows that the transformation method of conflicting rules is effective to reserve theinconsistent information. By considering the difference of recommendation problems and theexperience of decision makers, the utility analysis method can help recommendation systems to combine the results from multiple information sources and make better decisions for thepersonalized recommendation problem. In addition, integrating online reviews into therecommendation system is capable of predicting customers' true feelings and provides a novel ideato satisfy customers by product recommendation.(3) This paper investigates the integration of online price promotion and productrecommendation which helps companies maximize promotion profits by establishing attractive pricediscounts and recommending right products. Suppose a company is conducting promotion activitiesto maximize profits, this paper proposes probabilistic methods to calculate the promotion profits ofthe promoted product, and its complements, substitutes and value-independent products. We proposea profit-maximizing model and a Genetic Algorithm (GA) based solution methodology to optimizethe promotion discount and recommended products.Our study shows that companies should recommend right products while optimizing the pricediscount of the promoted products. The profits would be compromised if they separate the discountoptimization and product recommendation. The proposed model can drive ideal customer trafficwhile avoiding profit losses caused by the price discount, and maximize company profits byrecommending right products associated with the price discount.(4) This paper investigates the personalized price strategy from the perspective of customizedbundling which derives more purchases from customers and maximizes profits for companies.Employing "Buy more, Save more" as the promotion theme, we offer companies an online dynamicbundle pricing (ODBP) model to satisfy customer preferences, enhance customer savings andmaximize company profits. The ODBP model employs the nonlinear mixed-integer programmingmethod to simulate online customer behaviors and calculate personalized prices by trackingcustomer's multi-stage purchase decision. A heuristic method is designed to solve the ODBP modelquickly and to meet the real-time requirement of e-commerce context.Our study shows that the ODBP model can provide a smart description of customer's onlinepurchase behavior. The personalized price strategy starts with an insight into customer motivation tobuy more products, and ends in a stream of company's profit enhancement. The heuristic-basedsolution procedure is capable of achieving a near-optimal bundle price with negligible computationtime. Our study offers a novel inspiration for companies to conduct real-time personalized pricestrategy.The experiments based on UCI database and Amazon.com's recommendation practices showthat (1) the proposed model can efficiently discover customer's preference from the massive data;(2)the combination of the recommendation decisions from multiple information sources can optimizethe recommended products to satisfy customers;(3) the integration of product recommendation andpromotion strategy can enhance promotion profits; and (4) the integration of recommendation andprice strategy can attract more customers to buy more products at the personalized price andmaximize company profits.This paper extends the research directions of the personalized marketing strategy, enriches themethodology of the personalized modeling, and provides systematic guidelines to designpersonalized marketing strategies.