Big data multi-access selection method based on submodular function

Abstract

The invention provides a big data multi-access selection method based on a submodular function. The method comprises the following steps of: 1, collecting a plurality of features of an existing storein a to-be-selected area to construct a sample set, wherein the features can influence the site selection quality; 2, training an evaluation model of a plurality of base learners by using the sample set, and combining the plurality of base learners into a strong learner, namely a final site selection model, by using a bagging integration method; 3, respectively inputting the features of a plurality of candidate stores in the to-be-selected area into the site selection model, and selecting the candidate store corresponding to a maximum output value as a target store; and 4, removing the selected target store and other candidate stores within the influence radius of the target store, and repeating the step 3 to obtain a next target store. According to the method, the site selection model with a wider coverage range is constructed, and finally, a scientific, standardized and high-sustainability site selection strategy is realized.

Description

technical field [0001] The invention relates to the fields of machine learning and big data, in particular to a method for selecting multiple locations of big data based on a submodular function. Background technique [0002] With the development of social economy, the chain industry is in the ascendant. Whether in the catering, retail or hotel industry, chain operations are increasingly active. Site selection is a very important thing for a chain enterprise, and may even be an important link that determines the success or failure of the enterprise, because many other factors that affect the operation of the enterprise, such as marketing, personnel services, prices, etc., can be adjusted according to the current situation. Faster changes to better adapt to the current development situation, and once the site selection is completed, the location cannot be easily changed, and at the same time, the customer groups covered by the store itself are mostly determined. Therefore, ...

AI Technical Summary

Problems solved by technology

This solution is undoubtedly much better than the previous method of purely empirical addressing, but it also has some inherent shortcomings. First, the data is collected manually, which takes a long period, and the comprehensiveness and accuracy of the data are not very high. Difficult to quickly copy to the location of other stores

The second is the inherent problem of chain store location selection. The problem of mutual interference caused by multiple stores. When the number of stores is large, there will inevitably be overlapping influences between each other, and the income of a single store will decrease. be negatively affected

This method uses logistic regression to judge whether the area is suitable for a certain type of store opening, and sets all areas with this type as 1, ignoring the fact that stores in many industries are blooming everywhere, and the results of logistic regression are difficult to compare in multiple Choose a better location in a suitable preselected area

[0005] After the existing site selection system is established, only a single location can be selected as a pre-selected area through the model each time

If you want to open multiple stores, you must run the model multiple times or select the area with the highest effect as the location, without considering the impact of competition with existing stores

Therefore, this type of site selection method cannot meet the requirements of chain enterprises to cover as large a range of stores as possible.

The problem of trying to cover a larger area with limited stores is an NP-hard problem

Images