"Wordle" Distribution Prediction and difficulty Classification Prediction based on Deep Learning

Abstract

This project mainly focuses on the study of word guessing games, analyzing factors such as the popularity of the game, the playability of the game itself, the difficulty of the game itself, and the game effects of the participants. Firstly, analyze the useful attributes from the words and whether these attributes have an impact on the proportion of game score results reported on that day. Secondly, clustering models are used to automatically partition words, and the difficulty is divided into two levels: simple and difficult. The machine learning xgboost model is used to evaluate the model through confusion matrix, accuracy, recall, and F1 value. The study found that the total number of reports is consistent with the trend of the number of reports in hardcore mode. The popularity rapidly increased in the early stage and gradually decreased in the later stage, which means that the loyalty of players has basically been washed away, and the remaining are mostly people with great potential to continue playing.