Optimizing Facial Expression and Head Dynamics Data Processing to Enhance Depression Detection with Cutting-Edge AI Models

This dataset contains facial behavior and head movement data used to predict depression episodes. It includes features such as Action Units (AUs), head movement metrics, and other facial structure indicators.

---

Dataset Structure

The dataset consists of the following columns:

Column	Description
AU24	Action Unit 24 activation (Lip Pressor)
Eye_Open_Avg	Average eye openness
Facial_Structure	Calculated facial structure value based on geometric features
AU23	Action Unit 23 activation (Lip Tightener)
AU01	Action Unit 01 activation (Inner Brow Raiser)
AU07	Action Unit 07 activation (Lid Tightener)
AU02	Action Unit 02 activation (Outer Brow Raiser)
AU_Smile	Activation value of smile-related Action Unit
AU10	Action Unit 10 activation (Upper Lip Raiser)
AU14	Action Unit 14 activation (Dimpler)
Head_Movement	Head movement value (rotation and translation)
AU_Sad	Activation value of sadness-related Action Unit
datetime	Date and time of data collection
depression_episode	Binary label: 0 (No episode), 1 (Depression episode)
pid	Participant ID (e.g., P08)

---

Sample Data

Here is a sample of the dataset:

AU24	Eye_Open_Avg	Facial_Structure	AU23	AU01	AU07	AU02	AU_Smile	AU10	AU14	Head_Movement	AU_Sad	datetime	depression_episode	pid
-11.351092	0.973326	0.567567	-2.459438	-11.097750	18.764470	-13.701491	1.797318	-7.621264	-8.416746	0.093635	0.861408	2022-07-21 04:46	0	P08
-12.812418	0.824679	0.584101	-2.705608	-8.061482	13.727324	-12.742016	1.416385	-0.372542	-13.412502	0.094856	0.250375	2022-07-21 04:46	0	P08
-9.396528	0.905175	0.609942	0.163664	-5.724306	22.029821	-9.810085	1.645429	-4.372520	-3.623914	0.095847	2.480238	2022-07-21 04:46	0	P08

Universal Model Accuracy

Overall Evaluation of the KNN Model

The overall evaluation of the model based on the following metrics:

Metric	Value
Accuracy	0.6563
Precision	0.6303
Recall	0.6999
F1 Score	0.6183
AUC	0.7240

Overall Evaluation of the XGBoost Model

The overall evaluation of the model based on the following metrics:

Metric	Value
Accuracy	0.8167
Precision	0.8224
Recall	0.8351
F1 Score	0.7751
AUC	0.9155

Overall Evaluation of the Decision Tree Model

The overall evaluation of the model based on the following metrics:

Metric	Value
Accuracy	0.8475
Precision	0.8138
Recall	0.8682
F1 Score	0.8206
AUC	0.9135

Overall Evaluation of the Naïve Bayes Model

The overall evaluation of the model based on the following metrics:

Metric	Value
Accuracy	0.7291
Precision	0.6599
Recall	0.9410
F1 Score	0.7339
AUC	0.8821

Hybrid Model Accuracy

Extra Trees Classifier + XGBoost + Logistic Regression sebagai Meta Classifier

Class	Precision	Recall	F1-Score	Support
0	0.88	0.87	0.88	34,022
1	0.86	0.88	0.87	31,131
Accuracy	0.87	0.87	0.87	65,153
Macro Avg	0.87	0.87	0.87	65,153
Weighted Avg	0.87	0.87	0.87	65,153

Usage

Clone this repository:
bash git clone https://github.com/SkylarkOff/Semicolon-Model.git