K-Means Clustering for the coping strategies of Brief COPE Questionnaire

This repository contains the code and results of a K-means clustering implementation to extract different groups of coping strategies that influence resilience.

Table of Contents

• Introduction
• Data
• Methodology
• Results

Introduction

The Brief COPE is a widely used self-report questionnaire that assesses coping strategies individuals use when faced with stress or challenging situations. The questionnaire consists of 28 items that measure 14 coping strategies Self-distraction, Denial, Substance Use, Behavioural disengagement, Emotional Support, Venting, Humour, Acceptance, Self-Blame, Religion, Active Coping, Use of Instrumental Support, Positive Reframing, and Planning.

Understanding coping strategies’ impact on psychological well-being is key to identifying strategies that may serve as resources for successful adaptation. Existing research has explored the relationship between coping styles and various mental health variables, such as resilience. Resilience might be seen as a personality trait—a positive, distinct feature of an individual that mitigates the negative effects of stress and minimizes episodes of depression.

In this project, we apply the k-means clustering algorithm to cluster the coping strategies. The goal is to identify distinct groups of coping strategies that influence resilience. To do this, we first use different feature selection methods to extract important strategies that influence resilience. Then, we employ k-means clustering to cluster these coping strategies. Finally, by comparing the obtained clusters, strategies that can improve resilience are introduced.

Data

We utilized a preexisting dataset provided by Konaszewski et al. (Konaszewski K, Niesiobędzka M, Surzykiewicz J. Resilience and mental health among juveniles: role of strategies for coping with stress. Health Qual Life Outcomes. 2021 Feb 18;19(1):58) https://doi.org/10.3886/E120001V1. They investigate the direct and indirect role of resilience in shaping the mental health of juveniles. The dataset includes resilience and 14 coping strategies.

Methodology

Feature Selection

Before applying k-means clustering, we employ various feature selection techniques to extract important coping strategies that significantly influence resilience. The selected coping strategies are then used as input for the clustering algorithm. Notebook

Clustering

Once the relevant coping strategies are identified, we utilize the k-means clustering algorithm to group them into distinct clusters based on their similarities. Notebook

Results

Summary of feature selection: From the correlation matrix of data, it was observed that the variables 'Self_distraction', 'Active_coping', 'Emotional_support', 'UIS', 'Behavioral_disengagement', 'Positive_reframing', 'Planning', 'Acceptance' have substantial correlation with "resilience".

To sum up, the main features that were particularly important in influencing resilience are: Active_coping, Planning , Emotional_support, Positive_reframing, Acceptance, Behavioral_disengagement, Humor and Self_blame.

K-means clustering: Firstly, by using the Elbow method, three clusters were chosen.

Then, k-means was conducted and the mean values of the selected coping strategies in each cluster were obtained as

Further, the means of resilience in each cluster are obtained as

Based on these results, the main attributes of each cluster are summarized as follows:

Cluster 1: This cluster includes juveniles with the most resilience. They had high average for Active_coping, Emotional_support, Acceptance, planning and Positive_reframing and low average in Behavioral_disengagement, Self_blame and Humor.

Cluster 2: juveniles with the moderate resilience. This group had moderate average in almost all features and high average for Active_coping and Acceptance.

Cluster 0: This group had the lowest value of resilience characteristic. Active_coping, Emotional_support, Acceptance, planning and Positive_reframing were minimum for these juveniles.