Abstract:In the development strategy, the company frequently raises the task of improving the level of material management, optimizes procurement strategies, and improves the level and quality of material procurement. However, the market environment is complex and the variables involved are artificially difficult to grasp. For enterprises, the unfair bid winning in the tender procurement process has great impact and harm. Therefore, it is very important to grasp the interval value of the reasonable bid price in advance. To predict a reasonable price range, the method of this paper is divided into two steps. Firstly, an SVM model is obtained by training all previous bidding process data of the same type, and then a reasonable winning value is estimated based on the tender price adaptation model of the bidding. Because we can use a smaller number of feature values, we cannot solve the optimization problem with SVM tuning. How to find the reasonable deviation value at the right and left ends of this reasonable bid-winning value, this paper has tried in two ways. The first one compares the prediction result of a large number of similar data with the actual value, and uses the maximum value as the reasonable deviation value after filtering; The second way, the use of hypothesis testing principle to verify the existence of a percentage of the number of bids around the reasonable bid value, the bid price is often within this offset number range. The second way is a better solution to the situation where the bid price gap is relatively large. Under the premise that the historical bid price is a reasonable bid price, the accuracy can reach more than 90%, and ensure that the results have considerable reference value. It can be concluded that using a learning algorithm to predict a reasonable bid price is feasible.