Stellar Anomaly Detector

This repository provides an anomaly detection model trained on data from the James Webb Space Telescope (JWST). The model uses an Isolation Forest algorithm to identify anomalies in various JWST data types, including timeseries, images, spectra, catalogs, cubes, and measurements. Key parameters were finely tuned using a Randomized Search for Hyperparameter Optimization, and the model incorporates advanced Feature Engineering, Data Normalization, and detailed Performance Metrics to ensure high accuracy and reliability.

This approach specifically focuses on detecting anomalies introduced by malicious actors, such as falsified or manipulated data that may have been injected or altered intentionally. It aims to identify and mitigate the negative impact of such malicious data before proceeding with further analysis or workflows.

Data Acquisition

To obtain the data sets needed for running the model, follow these steps:

1. Visit the MAST Portal.
2. Go to the "Advanced Search" section.
3. Under "Mission," select JWST.
4. Choose the desired "Product Type" (e.g., timeseries, image, spectrum, catalog, cube, measurements) to download the data set. You can select one or more product types based on your analysis needs.
5. Under "Observation Type," select science to filter for relevant data.

The datasets used to develop this model were also sourced from this platform.

Model Performance

The model's performance varies depending on the data type used. Below are the results obtained on a dataset with 5,000 entries for each data type.

1. All Product Types (timeseries, image, spectrum, catalog, cube, measurements):

• Accuracy: 87.69%
• False Positive Rate: 6.66%

# AI Model Output
{'Total Anomalies Added': 999, 'Correctly Detected Anomalies': 876, 'False Positives': 229, 'Accuracy': 0.8768768768768769, 'False Positive Rate': 0.06664726426076834, 'Best Params IF': {'contamination': 0.24958833021525914, 'max_features': 0.7, 'max_samples': 0.6, 'n_estimators': 500}}

2. Images Only:

• Accuracy: 92.39%
• False Positive Rate: 7.74%

# AI Model Output
{'Total Anomalies Added': 999, 'Correctly Detected Anomalies': 923, 'False Positives': 303, 'Accuracy': 0.923923923923924, 'False Positive Rate': 0.07743419371326347, 'Best Params IF': {'contamination': 0.24958833021525914, 'max_features': 0.7, 'max_samples': 0.6, 'n_estimators': 500}}

How to Use the Anomaly Detection Script

This script allows you to detect anomalies in JWST data using the pre-trained model provided in this repository. Follow the steps below to use the script effectively:

1. Download the Necessary Files:

• Ensure you have the sad_model.pkl file in your working directory. This file contains the pre-trained anomaly detection model.
• Download the sample data file MAST_allmix_100_with_anomaly_JWST.csv from this repository, or use your own JWST dataset. The dataset should contain the following columns: s_ra, s_dec, t_min, t_max, t_exptime, em_min, em_max.

2. Prepare Your Environment:

• Ensure you have Python installed on your system. You will need Python 3.7 or later. The project was developed and tested in a Python 3.11 environment. It is recommended to use Python 3.11 to ensure compatibility and avoid potential issues.
• Install the necessary Python libraries by running:

pip install -r requirements.txt

Run the script using the following command:

python sad.py

Understanding the Output

• The script will print the rows in your dataset identified as anomalies based on the model.
• Use this feedback to analyze unusual data points that could indicate significant deviations in your dataset.

Note: The sample dataset MAST_allmix_100_with_anomaly_JWST.csv is designed to help you test the model. It contains injected anomalies to simulate realistic detection scenarios. The data from the 100th record onward consists of entries with anomalies.

Feel free to adjust the script according to your needs, such as changing the dataset path or modifying the feature engineering steps based on your specific data characteristics.

Further Reading

To learn more about the significance of this project and its use in anomaly detection, you can refer to the following article:

Detecting Anomalies in James Webb Space Telescope Data: Why It Is Important for Science and Security

License

This project is licensed under the Apache License, Version 2.0.
You may obtain a copy of the License at: http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, this software is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License. For more details, refer to the LICENSE file in this repository.