Welcome to SAM’s documentation!
The SAM Python package is a collection of tools and functions mainly for sensor data or time series analysis, and by extension for smart asset management analysis. All functionality has been tested and documented. Using this ensures a generic approach, and will greatly speed up analysis.
The easiest way to install the package is using pip:
pip install sam # To install all optional dependencies: (such as pymongo, seaborn, tensorflow, etc.) pip install sam[all]
There are different optional dependencies for SAM, if you are unsure use pip install ‘sam[all]’ other options include plotting (just use the plotting functionality), data_science (all dependencies needed for a data scientist) and data_engineering (dependencies for data engineer).
Keep in mind that the sam package is updated frequently, and after a while, your local version may be out of date with the online documentation. To be sure, run the pip install -U sam command to install the latest version.
Below you can find a simple example on how to use one of our timeseries models. For more examples, check our example notebooks
from sam.datasets import load_rainbow_beach from sam.models import MLPTimeseriesRegressor from sam.feature_engineering import SimpleFeatureEngineer data = load_rainbow_beach() X, y = data, data["water_temperature"] # Easily create rolling and time features to be used by the model simple_features = SimpleFeatureEngineer( rolling_features=[ ("wave_height", "mean", 24), ("wave_height", "mean", 12), ], time_features=[ ("hour_of_day", "cyclical"), ], keep_original=False, ) # Define your model, see the docs for all parameters model = MLPTimeseriesRegressor( predict_ahead=(1, 2, 3), # Multiple predict aheads are possible quantiles=(0.025, 0.975), # Predict quantile bounds for anomaly detection feature_engineer=simple_features, epochs=20, ) model.fit(X, y)
A configuration file can be created as
.config. This configuration file only stores api credentials for weather api’s for now.
The configuration file should be placed in your working directory, and is parsed using the
Python3 configparser, and an example configuration is shown below:
[regenradar] url=https://rhdhv.lizard.net/api/v3/raster-aggregates/? user=user.name password=secret [openweathermap] apikey=secret
SAM aims to support the whole analysis from ingesting data to measuring and visualising model performance. The package is build up accordingly as can be seen in the Contents below.
Easily train a functional model on timeseries data with
Build features from time series using the
Automatically remove extreme values, flatlines and missing values with
Visualise the predicted quantiles using
Contributing / requesting features
Contributing works by forking the repository and creating a pull request if you want to incorporate your changes into the code. Also see the CONTRIBUTING.md file in the repository for more information.
We keep track of new features, bug reports and progress on the GitHub issues page.
- Data formats
- Data Sources
- Data Sets
- Feature Engineering
- Data Validation