Research on Wordle Puzzle Game based on trend prediction model based on data index

Abstract

World Puzzle is a word game that has exploded in popularity over the past year. For this kind of game phenomenon, this paper studies. First, the LSTM algorithm is used to describe the trend of their number and to fit a high-precision time model. At the same time, considering the rigor, this paper also establishes the ARIMA time series analysis model to predict the data again. On this basis, the data accuracy of the two models is compared, the model is selected, and five indexes such as the number of words and the number of meanings are defined. Second, Pearson correlation coefficient was used to investigate the relationship between the effects of these indicators on the percentage of reported scores in difficulty mode. Finally, BP iterative neural network was used to calculate the correlation between the indicators and the percentage of players attempting to solve puzzles each time, and GBDT machine learning method was used to quantify the relationship between the word difficulty attribute and the reported result distribution, so as to establish the correlation model between the word attribute and the percentage of passing rate. The results show that LSTM model is more suitable than ARIMA model to describe the trend of the number of reports over time. Meanwhile, when the word "EERIE" was inserted into the model, the average number of correct guesses was 5.399.