A Study on Analyzing the Direction of Wordle Puzzle Results Based on ARIMA Prediction and Random Forest Algorithm

Wordle is a popular puzzle currently offered daily by The New York Times, and it has gained popularity among players from various countries and regions in a short period of time. In this study, by analyzing and studying the data of Wordle puzzles provided by The New York Times, we predicted the number of reported results of Wordle in the next two months, and comprehensively summarized and analyzed the word characteristics. First, an ARIMA prediction model was established and the trend of the number of Wordle reported results was simulated. Second, four attributes of words were extracted: word frequency, initial letter frequency, letter repetition rate, and number of words, which were analyzed using multiple linear regression model, correlation analysis, and ADF test. Then, the word attribute data were collected and organized and analyzed using an iterative random forest model, which showed that the model fit was good and there was no overfitting. Next, the words were classified by difficulty using the RSR comprehensive evaluation method, and the classification results were trained and tested using the machine learning decision tree classification model, with an average accuracy of 85%. Finally, the obtained data of the word “ERNIE” was used to predict the difficulty, and the conclusion was “easy”. The number of attempts for each word was analyzed by S-W test and K-S test and found to be normally distributed. This study provides an in-depth understanding of the future development trend of Wordle and word characteristics, and offers new methods and ideas for related research.