Exploring a dataset on hard drive failure, and then building a predictive (classification) model with H2O. This is a basic model built for the purpose of testing how to implement the model with Plumber as an API. As the subsequent steps are not done in Markdown, they will not be placed here. All files are available on GitHub at: https://github.com/supermdat/hard_drive_failure

Step 08 of a multi-part forecasting of Chicago subway ridership. Focusing on the visualization of accuracy metrics from the various models used.

Step 07 of a multi-part forecasting of Chicago subway ridership. Focusing Keras LSTM models.

Step 06 of a multi-part forecasting of Chicago subway ridership. Focusing on calculating accuracy stats for all models to this point.

Step 05 of a multi-part forecasting of Chicago subway ridership. Focusing on H2O AutoML models.

Step 04 of a multi-part forecasting of Chicago subway ridership. Focusing on ARIMA and Prophet models.

Step 03 of a multi-part forecasting of Chicago subway ridership. Focusing on Random Forest and XGBTree models.

Step 02 of a multi-part forecasting of Chicago subway ridership. Focusing on setting up the modeling infrastructure.

Step 01 of a multi-part forecasting of Chicago subway ridership. Focusing on obtaining data and feature engineering.

This notebook compares two regex models classifying the outcomes of service calls for rat abatement (i.e., classifying the outcome as `rats_found`, or `no_rats_found`, or `unknown`). Each model is run on two texts, a “raw text” and a “cleaned text,” for a total of four comparisons. A sample of manually categorized text is also used for creating Confusion Matrix Statistics.

This is a work-in-process notebook that uses natural language processing (primarily Latent Dirichlet Allocation) to inform a regex model which classifies the outcome of service calls for rat abatement (i.e., classifying the outcome as `rats_found`, or `no_rats_found`, or `unknown`). The specific dataset used is the "newer" version: dc_311-2017-10-07.csv

This is a work-in-process notebook that uses natural language processing (primarily Latent Dirichlet Allocation) to classify the outcome of service calls for rat abatement (i.e., classifying the outcome as `rats_found`, or `no_rats_found`, or `unknown`). The specific dataset used is the "older" version: dc_311-2017-01-16.csv

Notebook for analyzing data on DC buses.