LSTM_predictions_tf | Use LSTM in tensorflow to predict outcomes from text data

LSTM_predictions_tf is a Jupyter Notebook library. LSTM_predictions_tf has no bugs, it has no vulnerabilities and it has low support. You can download it from GitHub.

Use LSTM in tensorflow to predict outcomes from text data. The goal of these models was to train an LSTM network (long short term memory network, an artificial netural network architecture) predict depression labels (based on the PHQ-8) from transcribed verbal data. Data comes from from interivews with around 700 participants; these interviews were intended to simulate a structured psychiatric interview. Prediction options include: depression level (continuous, low/med/high, or binary), level for each of 8 possible depression symptoms based on PHQ-8 (low/med/high, or binary). You may need to change the cost function in the code depending on the outcome chosen. Depression data is often unbalanced (more lower levels of depression), so options in code for 1) undersampling the majority class (lower levels of depression) and 2) cost-senstive learning. For regularization, there are various options for dropout and L1/L2 regaularization.