Multivariate Time Series Forecasting

Job ID: 36602010

Budget: $30 – $250 USD

We are seeking an experienced Machine Learning Engineer. The project involves predicting 19 dependent climate variables, including temperature and humidity, which are collected from a climate-controlled environment. The predictions will be based on 28 independent variables, with the aim of forecasting the dependent variables at various time horizons, ranging from 1 to 8 hours.
The ideal candidate will leverage their expertise in Long Short-Term Memory (LSTM) networks, or similar technologies, to handle this multivariate time-series forecasting task. The data set consists of over 500,000 observations gathered over a span of more than a year. The input to NN has to be editable enabling the change in the last observations before prediction. The project has to contain documentation on evaluation metrics and their interpretation.