Research on Influencing Factors of Mobile User Experience in Beijing
The rapid development of mobile communication technology has brought great convenience to people, and people are increasingly inseparable from the various conveniences brought by mobile communication technology. With the continuous construction of the network, the network coverage is becoming more and more perfect. Various mobile operators pay more and more attention to the customer's network experience, so as to further improve the network service quality.
Analysis of Problems and Ideas
2.1 The first question
According to Annex 1 and Annex 2, the main factors affecting customer satisfaction with voice service and Internet service are studied respectively, and the quantitative analysis and results of each factor's influence on customer rating are given. See Annex 5 for explanations of each field in Annexes 1 and 2.
The main thing here is to clean the data first, and then calculate the feature importance of the indicators. It should be noted that the feature importance of the one-hot encoded indicators needs to be combined for calculation, and then the decision tree can be used to draw a decision map of satisfaction.
The first four are the cleaned data. It should be noted that when quantitatively analyzing the indicators here, there is no need to set up a test set, because the important thing here is to analyze the fitting degree of the overall data after training.
2.2. The second question
combines the analysis of question 1, establishes a mathematical model for customer ratings based on relevant influencing factors for customer voice services and Internet services, and based on this, conducts prediction research on customer ratings in attachments 3 and 4, and predicts the results respectively. Fill in the two worksheets of Sheet1 "Voice" and Sheet2 "Internet" in result.xlsx, and upload them to the competition platform to explain the rationality of your predictions.
This question is the key to distinguish the award. This is a classic classification model. Generally, machine learning models can be used. The mainstream in the industry includes machine learning such as xgboost, random forest, decision tree, etc., or deep learning time series prediction such as lstm, sequential model, etc. After selecting the model , here we need to establish a test set and a training set when training the model, and compare the effect of the evaluation function with the error of the analysis result.
The final prediction gets the result of the model and uploads the prediction.
There are three ways to improve the accuracy:
1. Increase the sample size
2. Optimize the model parameters, which can be combined with the TPE algorithm, genetic algorithm, and particle swarm algorithm for parameter tuning;
3. Build a new feature combination.
Specifically, I will refer to the detailed code It will be analyzed in the explanation video,
2022 Asia Pacific Competition Question C Code Acquisition Document
Summary of 2022 MathorCup Question B
The video idea has been released
2022 mathorcup question B complete problem-solving code and full nanny tutorial_哔哩哔哩_bilibili