We built a solution for health care professionals to determine if a patient will return to the hospital within thirty days. This uses a feed-forward deep neural network (DNN), which trains a model to predict the 30-day re-admittance of a patient. The DNN is built using the TensorFlow machine learning library for Python 3.5. The DNN has hidden two layers. The first hidden layer has 50 nodes, while the second hidden layer has 20 nodes. After training the model, a separate Python script is able to load the pre-trained model at-will and use it to predict for a single patient's data.
We have created a webpage (joshriess.tech) for health care professionals to more easily access the predictive capacities of the DNN model. This webpage allows for input of a patient's data and quick prediction. In order to accomplish this, the HTML input is pre-processed by PHP and passed as arguments into the Python evaluation script, where it is further processed and validated before use with TensorFlow. The final prediction is passed to PHP along with an identification string. PHP then identifies the final result and notifies the user.
Note: joshriess.tech may be password-protected at this time.
All products are in a prototype stage.