A Study on the Development of Wordle Based on Machine Learning

This paper builds a Wordle Word Difficulty Assessment Model based on Machine Learning which achieves input 5-letter English words, can output the predicted the percentage of different tries times on the specified date, and the word difficulty level (difficult, average or easy). First, the Regressor Chain is applied to correlate a set of independent variables (percentage of different tries) and dependent variables (frequency of each letter, word frequency, lexical properties, number of common letter combinations, number of repeated letters, date) in a certain order. Then, based on the Decision Tree, build Wordle Word n-tries percentage Prediction Model. The final implementation: input 5-letter word, output n tries percentage. Second, Using the optimized KNN clustering algorithm with percentage of n tries as a variable, three difficulty levels -"difficult", "medium" and "average" were classified. Next, the KNN Regression Model was optimized using the Particle Swarm Optimization (PSO) to build Wordle Word Difficulty Assessment Model. This model can classify the difficulty level of any 5-letter English word. By this model, it can also provide a solution to better enhance the user experience and increase the popularity of the game. In a way, this scientific analysis of Wordle historical data, which can better guide the future innovation and ideas of this popular game.