Research Of Keyword Bidding System Based On Machine Learning

With rapid development of the internet in recent years, online advertising has alsoachieved its quick development. Currently, Search Bidding Advertising is considered tobe the main pattern of online advertising and also the most efficient marketing deviceprovided by the internet. It helps enterprises to search for potential customers, and topopularize and sell products with assistance of the search engine, bringing profits to theenterprises. Many enterprises are taking key-word advertising as an importante-commercial marketing method. Key-word optimization of search bidding advertisinghas become a hot yet difficult issue in the field of the present search bidding advertising,which consists of2aspects, ie. key-word choosing and bidding of the purchased orreleased key words. At present, bidding of the purchased search-bidding words ismainly operated manually, ie. to adjust bidding of the purchased key words according toexperience of human-beings, which no doubt will cost much time and energy. The maincontribution of the present research paper is as following:To build a commodity navigation website with a focusing creeper system,collecting commodity information in the famous e-commercial website and storing itinto the database.To predict conversion rate of key words by using the logistic regression model.To propose three controling methods, ie. each-day efficiency number, each-dayconsuming amount, and consuming amount per efficiency number, and to prove theefficiency of the methods by experiments.To propose computational formula of key words’ prices by using key-wordconversion rate, controling methods, etc, predicted by the model and to purchase keywords in Baidu and popularize the commodity navigation website, calculating prices ofthe key words by using key-word bidding system and bidding through API of Baidu.