| Abstract [eng] |
The object of this research is the game Cryptokitties. It is based on blockchain technology. In this game you can collect, buy, sell and breed kitties with unique properties. This game has become quite popular both among players who want to have fun and among those who invest and try to earn money. An attempt was made to construct a model that would estimate the market price of kitties. Since they have so many features, estimating the price of a kitten has become quite a challenge for the player. In this research, we used two machine learning methods - random forest and regularized regression algorithm to predict the price of a kitten. The data was taken from three different sources. In order to obtain better results, new variables were created and the text field vectorization method Sent2Vec was used. The random forest model has given the best results. The most important variables in the model were related to ether and US dollar exchange rates, frequencies of kitten attributes and their pairs, fancy types, and other characteristics. However, the forecasting results are not very accurate, so it would not be worth using the model to develop an investment strategy. The popularity of this game has declined recently, but there is an increasing focus on virtual collectibles based on blockchain technology. It is therefore worth pursuing further research in this area, as it is likely that a sufficiently precise model and a profitable investment strategy will allow them to be applied in the future. |
