Cryptocurrencies Price Prediction Using Weighted Memory Multi-channels

Forecasting cryptocurrency price using convolutional neural networks with weighted and attentive memory channels

“…Although (Zhang et al, 2020) developed a Weighted Memory Channels Regression (WMCR) model that learning channel-wise features to improve the forecasting accuracy of cryptocurrency prices, the WAMC model proposed in this paper differs from the previous WMCR model in the following aspects:…”

“…• We combine the GRU component with self-attention mechanism to establish the attentive memory in this paper while the previous work (Zhang et al, 2020) only uses LSTM layers to memorize important information and ignore the dependencies of different time slots.…”

“…• We also analyze the impact of the depth of hidden layers in the channel-wise weighting module on both training loss and validation loss. Moreover, in this module, we tackle the information loss due to the global average pooling in (Zhang et al, 2020) by replacing it with GRU.…”

“…• We extend our analysis by conducting various ablation studies to examine the significance of each module and how the impact of each module on forecasting errors at different training ratios while these experiments have not been conducted in (Zhang et al, 2020).…”