The project reported in this paper will aim to investigate the effect of trading volume data on the performance of deep learning models used for cryptocurrency market price prediction.

Among many other crypto coins, Ethereum (ETH) is the coin of choice in this paper, as it will allow a more flexible interpretation of results and comparison to other studies in the field.

Forecasting is made based on the closing price of ETH, and the aim is to predict the closing price for the following week.

The main focus is made on deep-learning-based hybrid models made of Gated Recurrent Units (GRU) and Long Short Term Memory (LSTM), in addition to a customised Transformer model. The performance of the above models was evaluated before and after applying the Volume of day trading as an additional parameter.

The study results show that Hybrid Models do not experience significant improvements after adding Volume. However, Transformer model performance was notably harmed by Volume. Overall, it is not suggested to use the Volume of day trading as an additional parameter when performing cryptocurrency price prediction using deep neural networks.

Presentation slides explaining the project are available here.

The original pdf of the study can be found here.

Also, all the code is available on Git Hub via this link.

Telegram bot that enables users to show closing price for next week for number of tokens. The predictions are made using Hybrid Deep Learning model (GRU-LSTM), data is automatically updated and new prediction is made every night

Example of interaction with the bot:

Use "/start" command to start interacting with the bot. In order to get list of all available coins use "/menu", from the list of buttons choose coin of your interest.

Next bot will provide you with prediction of closing price for today and next six days ->

Here is your prediction of BTC closing price for the next 7 days:

                        Fri     29873
                        Sat     30165
                        Sun     30030
                        Mon     30354
                        Tue     29797
                        Wed     30458
                        Thu     30336
                    

All the code is available on Git Hub via this link.

You can try the bot by writing “Prediction Bot” in the Telegram search bar. Or access it via link.

The initial idea for this telegram bot was a simple AI-powered comments section analyzer. However, in the progress of development, we also introduced two new features: ContentBoost - taking the name of the item as input and providing the user with a comprehensive description of how to properly present this item on the marketplace website, this can be used as inspiration for a photographer of designer, additionally, you can generate the background idea of your liking straight away inside the bot. Item presentation - another feature that is meant to provide the user with design ideas for the background of the item's photo in the shop.

We launched the bot in December 2023, the number of our users is approaching 500, with the total number of requests processed close to 2,000!

Try it yourself: @ContentBoost_bot