Research And Application On Optimum Bidding Strategy Of Ad Words

AdWords is the most widely used and fastest growing way of launching advertisement in internet marketing.It is also known as Keyword Auctions or Position Auctions.Because of the unique closeness and uncertainty of keyword auction mechanism,advertisers will encounter two main problems in participating in the bidding process: firstly,how to determine the bidding price of keywords to ensure that advertisers can maximize their profits under this price;Secondly,how to do the budget allocation for multiple keywords under conditions of budget constraints in order to ensure the maximization of total revenue.This thesis is mainly to solve these two problems,and to develop a reasonable bidding strategy for advertisers to ensure that advertisers in the keyword advertising process can obtain the highest economic benefits.The main work of this thesis is as follows:1.Design a Multiple-LSTM data prediction model for Ad Wors benefits.Because of the different revenues obtained by advertisers under different bidding prices,the formulation of an optimal bid price strategy can be achieved by accurately predicting the bidding revenue of Ad Words.The prediction of bidding revenue needs not only to consider the bid price but also to consider the recent bid performance.Based on the characteristics of LSTM to do effective processing of time series data,and in order to better learn the correlation between multiple time series data,a Multiple-LSTM neural network model was designed.This model takes bidding price and nearly five days' bidding performance as input characteristics,and it can effectively predict the return profit,and experiments confirm that the model has a higher accuracy than other prediction methods.2.Use the second-generation Non-dominated Sorting Genome Algorithm II(NSGA-II)to solve the problem of budget allocation in Ad Words bidding.In detail,a multi-objective optimization model was established with the aim of maximizing revenue and minimizing invalid clicks.Using NSGA-II to solve the modle due to the existing advantages of uniformly distributed and well-diversified Pareto solution sets,and The method and process of this solving are discussed in detail..The NSGA-II algorithm is also used to analyze the ten key words in the actual bidding.The results show that the proposed algorithm can effectively improve the distribution of advertising budget and increase the advertisement revenue compared with the traditional average budget allocation scheme.The proceeds will have an instructive role for advertisers in formulating advertising budget allocation strategies.3.Design and develop an assistant management system in which advertisers participate in Ad Words bidding.The overall architecture of the system is also designed,including the architecture design of data acquisition module,data analysis module and interactive display module which is described in detail.The optimal bidding strategy and the optimal budget allocation strategy are integrated into the related modules to achieve the complete development of the whole system.