| Advertising has the meaning of publicity.Its main form is the dissemination of information through the media.As the fastest-growing media in history,the Internet has the advantages of fast communication,clear target audience,flexible updating,and low cost compared with traditional media such as paper newspapers,radio,and television.This also makes the Internet rapidly become an indispensable part of modern society.The search engine platform that relies on Internet technology has also become very important in national life.An important part of Internet marketing is Internet advertising,among which paid search relies on the search engine platform and is also an important source of income for the search engine platform.The largest search engine platform in China is Baidu,and its Baidu marketing platform accounts for the largest share of the domestic Internet marketing market.The characteristics of small and medium-sized enterprises are their small scale.and the traditional marketing model puts more pressure on them.Low-cost,high-return.and high-value-added marketing services are what these companies desperately need.Relying on the huge traffic of Baidu search engine,Baidu Marketing provides enterprises with online marketing services with low investment and high returns.This,to a certain extent,addresses the needs of small and medium-sized enterprises.In the process of cooperating with Baidu marketing platform,the key issue that enterprises need to consider is the input-output ratio.Predicting various indicators of enterprises in Internet marketing activities has become an important basis and method to guide marketing work.In order to complete the forecasting work proposed above,this paper predicts the average click price of Baidu marketing platform by drawing on the time series forecasting method often used in quantitative investment.Since traditional time series analysis usually only uses linear models or nonlinear models for modeling,the model will lose some effective information when learning sample features.In order to reduce the loss of information,this paper uses the LSTM-ARIMA hybrid model to learn the test samples.The LSTM-ARIMA hybrid model is composed of the LSTM algorithm and the linear model ARIMA.First,the LSTM algorithm is used to learn the nonlinear information in the training samples,and then the differential autoregressive moving average ARIMA model is used to learn the linear information in the residual to finally complete the error correction.According to the final experimental results,it is concluded that the LSTM-ARIMA hybrid model performs better than the single model. |
