TOPSIS is a sorting method that is close to the ideal solution. It determines the ranking of the evaluated object according to distance between the evaluated object to "ideal solution" and distance between the evaluated object to "negative ideal solution". There are two shortcomings in traditional TOPSIS: first, the distance from the evaluated object to the "ideal solution" and to the "negative ideal solution" is added directly in the calculation of relative closeness, which is unreasonable to add the positive index and the negative index directly; the other is that the traditional TOPSIS method ignores the weight distribution problem of the distance between the evaluated object and "ideal solution" and "negative ideal solution". In view of the above two problems, it proposes to use the osculating value method to modify the traditional TOPSIS method to solve the combination problem of two distances to positive and negative ideal solutions. Firstly, the distance from the evaluated object to "ideal solution" and "negative ideal solution" is transformed into the same direction index through osculating value, which ensures that the two distances can be calculated. Then, normalize the two distances to ensure that they are in the same order of magnitude, and that the distance between the alternatives and the ideal solution is equivalent to the distance between the alternatives and the negative ideal solution. Then, according to the preference of the decision-maker, distribute the weight of the two normalized distances to ensure that the weight of the two distances distribution really reflects the will of the decision-maker. Finally, the feasibility and effectiveness of the improved method are verified by an empirical example.