Automobile service chain enterprise shop opening and site selecting method based on LightGBM model

Abstract

The invention discloses an automobile service chain enterprise shop opening and site selecting method based on a LightGBM model. According to the method, H3 coding is adopted to carry out sample expansion on each store sample, multi-level and multi-granularity feature construction is carried out on each expanded sample by taking a regional grid, a sample store and a city as granularities, and corresponding numeric feature data are collected according to the constructed features; and a data set is formed by the income generated by each expanded sample in the recent certain time period and the corresponding feature data to train the lightGBM model to obtain a shop opening site selection income prediction model. The feature data corresponding to the shop opening candidate points is input into the shop opening site selection income prediction model to perform income prediction. According to the scheme of the invention, sample expansion is carried out through H3 coding to overcome the overfitting problem caused by insufficient machine learning samples, so that the generalization ability of the trained model is high.

Description

technical field [0001] The invention relates to the field of application of machine learning methods, in particular to a method for selecting a location for an automobile service chain enterprise based on a LightGBM model. This method is mainly used to predict the revenue of candidate locations for opening a store, and to guide the site selection business personnel to select the location for opening a store. Background technique [0002] As the cost of data acquisition becomes lower and lower, location information, navigation data, population distribution, crowd portrait data, and competition information can all be obtained through certain channels at a controllable cost. How to obtain effective information from the city's massive information and tap its value to guide the location selection of stores has become particularly urgent. [0003] The current location selection method based on machine learning is usually based on the characteristics and sales of historically open...

AI Technical Summary

Problems solved by technology

[0004] The existing technology uses qualitative methods to classify the characteristics of candidate points for opening a store, especially the environmental characteristics (such as business districts) around the storefront. From the objective reality, some practical There is no difference in relevant characteristics between stores with different online income, and the subjective component of human experience is relatively large

However, only summarizing the population information of the arriving business district as a statistical feature will ignore the influence of a large number of spatial distribution characteristics around the candidate point of opening a store on the candidate point of opening a store

Finally, because the training of models related to machine learning algorithms requires a large number of samples, the actual situation is often due to insufficient sample size resulting in data overfitting and poor generalization ability of the model.

Images