Market makers provide liquidity to small investors. To optimally provide liquidity and deal with potential demand spikes, it is important to know the liquidity required in advance. This chapter describes an algorithm that attempts to predict the liquidity for a 4-hour period using previous transaction data, return data, sentiment data, and macroeconomic data. One of the challenges to consider is the daily and weekly seasonality of order data. The main algorithm is based on LSTM, with data being preprocessed using a wavelet-based low-pass filter. As alternatives, SARIMAX and TBATS algorithms are considered, as they can also deal with seasonality. Experiments have shown that while all three algorithms capture seasonality well, the LSTM-based algorithm outperforms the other two overall, mostly by dealing slightly better with unusual situations. In addition to that, an ensemble model was built that combines the three approaches using a ridge regression, boosting the performance further. Data for experiments was provided by Nuri, a market maker in Berlin, Germany.