A Random Forest-Based Word Difficulty Prediction Model

The aim of this paper is to analyse the relationship between the word of the day and the corresponding distribution of the number of attempts in the Wordle game and to give a prediction method for the proportional distribution of word attempts. Firstly, the paper preprocesses the data provided by Question C of the 2023 American Collegiate Mathematical Modelling Competition. By constructing a model, this paper quantifies the word information entropy and people's preference for choosing common letters. Considering the above features and the influence of previous attempts on the follow-up, this paper constructs a regression model to verify the correlation between the word composition features and the distribution of the number of attempts. Meanwhile, considering the subjectivity of feature selection, this paper constructs a random forest model for further analysis. Comparing the results of the model analysis, the random forest model fits better, and the proportion of word EERIE attempts from 1 to 7 is 0%, 1.97%, 15.99%, 36.31%, 29.83%, 13.34%, and 2.24% respectively . This paper provides a theoretical basis for predicting the number of attempts of the corresponding words, which helps Wordle to optimally adjust the lexicon.