Highway transportation price index is a measure of the fluctuation of highway transportation market, which has an important predictive function for China's highway transportation industry. The extreme learning machine (ELM) neural network model is used to predict the highway freight index quickly and at low cost. Using the Person correlation coefficient of Baidu index and highway freight index to determine the impact of component on the highway freight index, and then the ADF stationarity test and Johansen cointegration test were used to construct the input sequence, and finally the window scroll was used to optimize the input variables. The predicted value was output in the ELM model. The results show that the ELM model based on the rolling window improved average absolute percentage error and root mean square error compared with the single ELM model. The prediction results are consistent with the index fluctuations, which can provide a decision reference for the direction of the highway freight index.