/Boston_311

Primary LanguageJupyter Notebook

Boston311Model

This is a class that defines a model to predict the likelihood of a 311 request in Boston being closed, given a set of input features. The model is built using tensorflow, sklearn, and lifeline libraries.

Tutorial

The tutorial.ipynb notebook shows examples of how to install this class and use it to train various types of models to predict the outcomes for Boston 311 cases.

Installation

To install the package, you can use pip:

pip install git+https://github.com/mindfulcoder49/Boston_311.git

Class Attributes

  • model: The trained machine learning model used for prediction.
  • feature_columns: A list of feature column names used in the model.
  • feature_dict: A dictionary where keys are the feature column names and values are lists of all the possible values for that feature.
  • train_date_range: A dictionary with keys "start" and "end", containing the start and end dates (as datetime objects) for the training data.
  • predict_date_range: A dictionary with keys "start" and "end", containing the start and end dates (as datetime objects) for the prediction data.
  • scenario: A dictionary of scenarios that can be applied to the data during cleaning. The valid keys and values are described in the clean_data() method.

Class Methods

__init__(self, **kwargs)

Constructor method for the class. Initializes all the class attributes with values passed as keyword arguments.

load_data(self)

This method loads the data for training the model. It returns a Pandas DataFrame containing the data for the specified time period.

enhance_data(self, data)

This method enhances the data by adding new columns. Specifically, it calculates the survival time (time between open and close date) and event (whether the request has been closed or not) columns. It also extracts the ward number from the ward column and converts the survival time to hours.

clean_data(self, data)

This method drops any columns not in feature_columns, creates the feature_dict, and one-hot encodes the training data.

apply_scenario(self, data)

This method applies scenarios to the data, if specified in the scenario attribute. The valid keys and values for the scenario dictionary are:

  • dropColumnValues: A dictionary of column names and lists of values to drop.
  • keepColumnValues: A dictionary of column names and lists of values to keep, all others being dropped.
  • dropOpen: Drop all open cases after a certain date.
  • survivalTimeMin: Drop all closed cases where survival time is less than a given number of seconds.
  • survivalTimeMax: Drop all closed cases where survival time is more than a given number of seconds.
  • survivalTimeFill: Fill in the survival_time for open cases based on a supplied end date as a string.

train_model(self, data)

This method trains the machine learning model using the cleaned data. It splits the data into training and testing sets, scales the features, and trains the model using logistic regression.

predict(self, data)

This method makes predictions on new data using the trained model. It returns a numpy array of predicted values.