Given ground truth data and forecasts, assess the model's performance by computing using time series regression metrics. Optionally v

Make time series datasets. Each example is a window of the last T_x data points and label is data point 1 day into the future.

Make time series datasets. Each example is a window of the last T_x data points and label is data point 1 day into the future.

Fits an RNN forecasting model :param dataset: A Pandas DataFrame with feature columns and a Consumption column

Load preprocessed dataset and return training and test sets. :param cfg: Project config :return: DataFrames for training and test sets

Load all entries for water consumption and combine into a single dataframe :param cfg: project config :param save_raw_df: Flag indicating

Transform raw water demand data into a time series dataset ready to be fed into a model. :param cfg: project config :param save_raw_df: F

Train a single model. Use the passed hyperparameters if possible; otherwise, use those in config. :param cfg: Project config :param hpara

Calculates estimates for daily water consumption based on provided historical data. Assumes each client consumes water at a uniform rate over

Runs k-prototypes clustering algorithm on preprocessed dataset :param k: Desired number of clusters :param save_centroids: Boolean indica

Perform a nested cross-validation with day-forward chaining. Results are saved in CSV format. :param cfg: project config :param dataset:

Evaluates performance of RNN model on test set :param train_set: A Pandas DataFrame with feature columns and a Consumption column

Fits an RNN forecasting model :param dataset: A Pandas DataFrame with feature columns and a Consumption column

Generate a forecast for a certain number of days :param days: Length of forecast :param cfg: Project config :param model: Model objec

Create a forecast for the test set. Note that this is different than obtaining predictions for the test set. The model makes a predic

Plot Prophet model's forecast using the Prophet API, including changepoints :param prophet_model: Fitted Prophet model :param prophet_pre

Train a model :param cfg: Project config :param model_def: Class definition of model to train :param hparams: A dict of hyperparamete

Conducts a Bayesian hyperparameter optimization, given the parameter ranges and selected model :param cfg: Project config :return: Dict o

Produces a box plot for all features in the dataset :param client_df: A DataFrame indexed by client identifier :param save_fig: Flag indi

Produces a correlation matrix for a dataset :param dataset: A DataFrame :save_fig: Flag indicating whether to save the figure

Decompose model into its trend, holiday, weekly and yearly components. Generate a plot and save parameters. Creates a new directory w

Abstract method for TensorFlow model definition

Abstract method for TensorFlow model definition

Create a forecast for the test set. Note that this is different than obtaining predictions for the test set. The model makes a predic

Create a forecast for the test set. Note that this is different than obtaining predictions for the test set. The model makes a predic

Create a forecast for the test set. Note that this is different than obtaining predictions for the test set. The model makes a predic

Given a preprocessed dataset, get the most recent factual example :param dataset: A DataFrame representing a preprocessed dataset

Loads the model from disk :param model_path: Path to saved model

Loads all raw water demand CSVs available and merges it into one dataset, keeping the latest consumption records for each client if readings

Plot all 2D hyperparameter comparisons from the logs of a Bayesian hyperparameter optimization. :param model_name: Name of the model :par

Plot model's predictions on training and test sets, along with key performance metrics. :param forecast_df: DataFrame consisting of predicted

Plot Prophet model's forecast components. This plot visualizes trend, yearly seasonality, weekly seasonality, holiday effects :param prop

Create a DataFrame, indexed by client, that contains client attributes as of a given date. Computes clients' monthly consumption over the las

Preprocess a new raw data file and merge it with preexisting preprocessed data. :param cfg: Project config :param save_df: Flag indicatin

Saves the model to disk :param save_dir: Directory in which to save the model

Perform Silhouette Analysis to determine the optimal value for k. For each value of k, run k-prototypes and calculate the average Silhouette

Train all models that have available definitions in this project :param cfg: Project config :param save_models: Flag indicating whether t

Run a training experiment :param cfg: Project config :param experiment: String defining which experiment to run :param save_model: Fl

Obtain general statistics for features in the client dataset and create a summary figure :param client_df: A DataFrame indexed by client iden

Plot average silhouette score for all samples at different values of k. Use this to determine optimal number of clusters (k). The optimal k i

Produces a violin plot for consumption by client in the most recent month stratified by rate class :param client_df: A DataFrame indexed by c

Evaluates performance of ARIMA model on test set :param train_set: A Pandas DataFrame with 2 columns: Date and Consumption :p

Evaluates performance of SARIMAX model on test set :param train_set: A Pandas DataFrame with 2 columns: Date and Consumption

Evaluates performance of scikit-learn model on test set :param train_set: A Pandas DataFrame with feature columns and a Consumption c

Evaluates performance of Prophet model on test set :param train_set: A Pandas DataFrame with 2 columns: Date and Consumption

Abstract method for model evaluation

Fits an ARIMA forecasting model :param dataset: A Pandas DataFrame with 2 columns: Date and Consumption

Fits a SARIMAX forecasting model :param dataset: A Pandas DataFrame with 2 columns: Date and Consumption

Fits a Prophet forecasting model :param dataset: A Pandas DataFrame with 2 columns: Date and Consumption

Abstract method for model fitting

Create a forecast for the test set. Note that this is different than obtaining predictions for the test set. The model makes a predic

Abstract method for forecasting with the model

Given a preprocessed dataset, get the most recent factual example :param dataset: A DataFrame representing a preprocessed dataset

Loads the model from disk :param model_path: Path to saved model

Loads the model from disk :param model_path: Path to saved model

Loads the model from disk :param model_path: Path to saved model

Loads the model from disk :param model_path: Path to saved model

Abstract method for restoring the model from persistent storage

Produces a series of data visualizations for client data and preprocessed consumption data. :param preprocessed_path: Path of preprocessed da

Saves the model to disk :param save_dir: Directory in which to save the model

Saves the model to disk :param save_dir: Directory in which to save the model

Saves the model to disk :param save_dir: Directory in which to save the model

Saves the model to disk :param save_dir: Directory in which to save the model

Abstract method for serializing the model