RankSNN: A Ranking Method to Integrated Evaluation

Evaluations are frequent, significant cognitive activities in human society. People make comprehensive judgments and rank the evaluated objects based on multiple factors, including relevant information and alternative data. Existing integrated evaluation methods assign weights based on expert opinions or data dispersion and provide evaluated rankings of objects through simple mapping. These methods are easy to implement, but the models produce biased results when the differences of all features in the evaluated objects are not significant enough. In order to reduce this error, we borrow the ideas from ranking methods and train the model based on ordering relations. We re-formulate the evaluation problem, construct a ranking model, and then design a new method named Ranking by Siamese Neural Network (RankSNN) adaptive to the model. This method uses ordering relations of sample pairs as labels to train the model, and measure the differences between samples by Siamese structure to get rankings. In this way, a trained scoring function can accurately correspond to the ordering relations among the sample pairs by calculating evaluation values and obtaining the best evaluation ranking results. In order to verify the validity and rationality of RankSNN, we conducted experiments on the public dataset of learning to rank and compared our results with the baseline ranking method. Then, we use RankSNN in integrated evaluation problems related to China’s urban business environment for the purpose of conducting empirical analyses, and the results of our method are superior to existing methods.