The traditional CPI forecast model uses the data mainly from the government bureau of statistics, because the government statistics data have low noise and are regular, which makes the forecasting model predict better in the period of little CPI change, but its effect is poor in the period of CPI turning point. However, network search data as a new type of data structure, which is applied to the forecast of economic and social problems, and its realtime available features, can predict in advance the turning point of the trend. Therefore, this paper adds the network search data to the traditional CPI forecasting model, and analyzes whether increased network search behavior can promote the prediction effect of CPI, especially for the CPI turning point period. The analysis results show that in the period of the turning point of CPI, the network search data can be added to the traditional model, and the model’s prediction effect can be promoted.