COMPARISON OF SUPERVISED LEARNING WITH UNSUPERVISED LEARNING ALGORITHMS IN DEPRESSIVE POST DETECTION

According to the latest WHO data published in 2017 Suicide Deaths in Kazakhstan reached 4,855 or 3.55 % of total deaths. The age adjusted Death Rate is 27.74 per 100,000 ofpopulation ranks Kazakhstan #4 in the world. This article shows the comparison of supervised and unsupervised machine learning algorithms, for detecting of depressive content in posts in social networks with emphasis on hopelessness and psych-ache for semantic analysis as the key reasons for suicide. Suicide is not an impulsive act and preparation for suicide can last about a year, during which a person will show signs of his condition in our case posting depressive content on his social network profile. This algorithm helps in detections of depressive content which can cause suicide, to help founded persons reach confident help from psychologists of national suicide preventing center in Kazakhstan. Obtaining highest result for 95 % of f 1-score for Random Forest(supervised) with tf-idf vectorization model, in conclusion of comparison we may say that K-means (Unsupervised) using tf-idf shows impressive results, which is only 4 % lower in f 1-score and precision.

1. World Health Organization. Preventing suicide. A resource for counsellors. Geneva 2006.

2. Oksana Lysenko. “The number of suicides among children in Kazakhstan continues is growing”, [http://www.zakon.kz/4524024-kolichestvo-suicidov-sredi-detejj-v.html]

3. Klonsky E. D., May A. M. “Assessing Motivations for Suicide Attempts: Development and psychometric properties of the Inventory of Motivations for Suicide Attempts (IMSA).” Suicide and Life-Threatening Behavior, October 2013: 1-3.

4. Mukhtarkhanuly D., Abishev A. “Suicidal Post Detection in Social Networks using NLP.” Natural Sciences Publishing, August 2018.

5. Marouane Birjali, Abderrahim Beni-Hssane, and Mohammed Erritali. “Prediction of Suicidal Ideation in Twitter Data using Machine Learning algorithms.” International Arab Conference on Information Technology (ACIT’2016), 2016: 1-5.

6. Hardik A. Patel, Cheng-Yuan Hsieh - Knowledge Systems Institute. “Early Detection of Suicide Using Big-Data Analytics in Real Time.” Journal of Visual Languages and Sentient Systems, 2016:1

7. Liu, Tong, Qijin Cheng, Christopher M. M. Homan, and Vincent M. B. Silenzio. “Learning from various labeling strategies for suicide-related messages on social media: An experimental study.” arXiv: 1701.08796v1 [cs.LG] 17, no. 01 (Jan 2017): 1-8.

8. Receiver Operating Characteristic with cross-validation [http://scikit-learn.org/stable/auto_examples/model_selection/plot_roc_crossval.html#sphx-glr-auto-examples-model-selection-plot-roc-crossval-py]

9. https://github.com/DaniyarML/Publications/tree/master/Supervised_vs_Unsupervised