The Research And Application Of Some Modeling And Optimization Problems Of Advertising Planning

Advertising, as a method of spreading information, is an important economic phe-nomenon and activity in modern commercial society. Because of its importance, advertis-ing campaign attracts more and more attentions. Many companies increase their invest-ment on advertising. Therefore, the capital input in advertising market increases year by year. How to acquire the advertising effect as well as possible becomes the problem that the decision makers are most concerned about.Taking a real project of a financial institute in Japan as the background, research and application of some modeling and optimization problems of advertising planning are carried on. Furthermore, based on the research results, advertising optimization system is designed and implemented. The main contents and works of this dissertation are summa-rized as follows:1). Based on much literature review, a survey on current research status of advertis-ing budgeting, budget allocation and advertising media selection is made.2). An optimization model for advertising budgeting is developed for the advertising campaign of which the objective is to use the minimum capital investment to achieve the desired advertising effect. The advertising budgets of each medium are selected as the decision variables. The objective function is the total advertising budget of the year. The number of response is selected as the main constraints. In order to evaluate the advertising effect, response models depicting the relation between the number of response and adver-tising budget are constructed. Because the number of response is selected as the evalua-tion criterion of advertising effect, developed optimization model is especially suitable for the advertising campaign of service industry. In order to deal with the duplicate effect of media-mixed advertising, inclusion-exclusion principle based media-mixed advertising’s total effect calculating model and LS-SVM regression based media-mixed advertising’s total effect calculating model are constructed respectively for two case where available data are advertising effects of each medium or available data are advertising effects of each medium and duplicate effects of them. Based on the media-mixed advertising’s total effect calculating model, the advertising budgeting and budget allocation between media are completed simultaneously. According to the characteristics of optimization problem and the decision variables, the real-coded genetic algorithms are applied to find optimal solutions.3). An optimization model for advertising budget allocation over planning periods is developed. The decision variables are the monthly budgets in planning horizon. The objective function is the maximal number of response. Annual budget and ROI of each month are selected as the constraints. Taking the carry-over effect into account, a LS-SVM regression based response model is constructed to evaluate the effect of advertising budget allocation’s result. According to the characteristics of decision variables and the forms of response models, the real-coded genetic algorithms are applied to find optimal solutions. In order to guarantee the feasibility of the solutions in the initial population, a novel initial population generating method pertaining to the advertising budget allocation problems is proposed.4). Existed evaluation criteria of the effect of media-mixed advertising are almost based on the psychological analysis of customers or the qualitative criteria. Though there are some quantitative criteria, they are just the criteria to evaluate the effect of media ve-hicles but not the effect of advertising. In advertising campaign, the evaluation criterion having relation with the economic benefit is required. Considering these situations, an au-dience rating data based model is developed to predict the number of audience who will respond to an advertisement. In addition, the estimation method of the parameter of pro-posed model are introduced. Proposed predicting model will be used in the optimization of media selection.5). Existed media selection methods set vehicles’s effects based on decision makers’ subjective experience. Furthermore, they can not deal with the problem that the unit costs decrease with the increasing appearing times. In order to solve these problems, an objective data based optimization model for media selection is proposed. The model proposed to predict the number of audience who will respond to an advertisement is used as the evaluating criterion of media selection result. Considering that available media vehicles become more and more but not all the of them have effect on target customers, based on the information gain, the demographic character based media selection method is improved to pre-select the media vehicles, which will reduce the solution space. Since the proposed optimization model for media selection is a nonseparable nonlinear integer programming problem for which there is no special algorithm, the genetic algorithm is adopted. An encoding method pertaining to optimization problem of media selection is proposed. Taking the characteristics of the chromosomes into account, the crossover point setting method is given.6). Based on the research of the optimization of advertising and the requirement of customer, the advertising optimization system is designed and implemented. The design scheme and framework of the system are given in Chapter6. The design of database and the functions of important modules are introduced by pictures of interfaces.Finally, future directions for research of the modeling and optimization of advertising planning are discussed after summarizing the whole work in this dissertation.