Conflicting-Competency

This branch is contributed by Kartik Satish Gawande (MT2023045).

Directory Details

classification/16th to 30th participant/

This directory contains earlier classification attempts using Random Forest and KNN on data from the 16th to 30th participants. These attempts were conducted:

• Participant Agnostically
• Participant Non-Agnostically

Key File:

• Results.jpg - Contains the results of the training.

---

pre-processing/Kartik/Video timestamps/

This directory includes Python scripts used to output the finalized training timestamps.

Key Files:

• Synced_Times_All.csv - Contains all the offsets for each participant's video timestamps to synchronize them.
• Other intermediate files are also present, which were used to derive the final synced file.

---

pre-processing/Kartik/Video timestamps/stamps/

This subdirectory contains all the interaction timestamps for each participant. These timestamps were:

• Semi-automatically extracted using script.py from the pre-processing/Kartik/Video timestamps directory.

---

pre-processing/Kartik/

This directory contains Python notebooks that were used to calculate the start and end timestamps for each activity type and file for each participant. The activity types are:

• 1, 2, and 3 (representing different activity types)
• Files are specified as 1, 2, 3, and 4 (representing the first, second, third, and fourth files respectively).

Key File:

• exp_timestamps_refined.csv - The final file containing timestamps synced with dataset timestamps. These are the final training timestamps.

---

Note: All timestamps in exp_timestamps_refined.csv are synced with the dataset timestamps.

kedro/

This directory has the entire codebase (including from all the other directories) shifted to kedro framework for better handling of the workflow. Please use this folder using kedro.

pip install kedro

Kedro Project Structure Overview

Directories and Their Contents

• notebooks: Contains all the scripts and Jupyter notebooks used across different parts of the project.
• src/kartik_egg/pipelines: Each pipeline directory has two main files:
  • nodes.py: Contains all functions for data manipulation and model training.
  • pipeline.py: Specifies the execution order of nodes and the data flow (inputs and outputs) between them.
    • Note: Many pipeline variables are instantiated here, but only one can be used at a time when running a pipeline.

Running a Pipeline

To run a specific pipeline, use the following command:

kedro run --pipeline=<pipeline_name>

Configuration Files in conf/base/

• The directory contains all the configurations used by the nodes during runtime.
  • parameters_training.yml: Directly specify model parameters like the number of estimators for the RandomForest model.
  • parameters_preprocessing.yml: Describes how files in the dataset are labeled and other preprocessing parameters.

Data Directory Structure in data/

• Well-organized according to their role in the pipeline workflows:
  • 00_scattered: Initial dataset files that are totally scattered.
  • 01_raw: Dataset files appended into a single file.
  • 02_intermediate/video_data: Intermediate data that cannot be directly used in training.
  • 03_feature: Contains final training stamps with mentioned label columns.
  • 04_primary: Holds the training-ready dataset.
  • 05_model_input: Training and testing split files.
  • 06_models: Saved models.
  • 08_reporting: Necessary reports from each trained model in JSON format.

Additional Resources

• For a better understanding of Kedro, please watch this 5-minute video: