Word game prediction and difficulty classification based on machine learning

Abstract

This article focuses on the Wordle game, a word puzzle game that has become extremely popular on social media. Based on the research objectives, three models have been proposed to solve the problem: a model for predicting the number of players, a model for predicting the number of attempts, and a model for classifying word difficulty. For the problem of predicting the number of players, the (Autoregressive Integrated Moving Average-Support Vector Machine) ARIMA-SVM model was proposed. The results showed that the ARIMA-SVM model performed about 17% better than the ARIMA model. To address the problem of predicting the number of attempts, the Random Forest Regression model was proposed. In regard to the problem of classifying word difficulty, the GA-FCM clustering model was proposed. Finally, this paper contributes to the difficulty classification of words based on different word attributes, and the experimental results showed that (Genetic Algorithm-Fuzzy C-Means) GA-FCM performed significantly better than the FCM model in all indicators.