Detecting and Terminating the Mental Disorders by Enhanced Multiple Classification in Social Network

Detecting and Terminating the Mental Disorders by Enhanced Multiple Classification in Social Network

1. Yoga

   (AP/IT)

   Department of Information Technology, Nandha College of Technology, Erode,India.

2. S. Akash Kumar, Biddu Mohan,

B. Gowri Sankar, N. NaveenKumar

(UG Scholar), Department of Information Technology, Nandha College of Technology,

Erode,India.

Abstract: The rates of diagnosing depression and mental illness during the last few decapods, a number of cases prevail unheard-of. Symptoms linked to mental illness are detectable on Twitter, Facebook and web forums and automatic methods are more and more able to locate inactivity and other mental disease. In this paper, latest studies that planned to detect depression and mental illness by the use of social media are surveyed. Mentally ill users have already been pointed out the use of screening surveys, their community distribution of analysis on twitter, or by their membership in online forums, and that they were detectable originating at regulate users by patterns in their language and online activity. Various automated detection methods can help to detect depressed people using social media. In addition a number of authors experience that various Social Networking Sites activities may be linked to low self-confidence, particularly in young people and adolescents.

Keywords: Depression, Mental state, Social Media, Machine Learning, Deep learning.

1. INTRODUCTION

   Elvis Saravia et al.,mention in his paper that the people that is suffering from mental illness usually wants to alone, and because of that they seek Social media as a platform to share their feelings and illness [1].Maryam Mohammed Aldarwish, Hafiz Farooq Ahmed put in the picture that the comprehensive use of social media could provide opportunities to help detect the depression which is undiagnosed. From the activities of the user in social media, we get the actions and behavior of mentally depressed patients and the way of thinking[2]. Social networking sites such as Twitter and Facebook modified the ways that other people describe their opinions, be in contact with others and share their experience. This result in the continuous flow of large amount of data containing important information related to sentiments and opinions of peoples.

   Being so regular Social Media platforms make massive quantities of compilations. Priya Nambisan et al., mention that studies related to depression suggests that persistent thoughts and deliberating behavior are the two main symptom characteristics [3].Social media platforms are normally consistent and provide users to access the internet. Traversely Conventional media requires a lot of time for

   compilation of information for publication. Even though Social Media is generated in real time.Various researches have been done in this field and the rate of depressed users is increasing day by day and this also sometimes led to suicidality. According to social media statistics the total Worldwide population is 7.6 billion and the internet has 3.5 billion users and among that 3.03 billion active social media users. Facebook has its own 2.072 billon users and it adds 500,000 users every day and Twitter has 330 million users. Facebook Messenger and Whatsapp handle 60 million messages a day.

   Studies up to now either consider to predict how, using Social networking sites correlates with mental illness or attempted to detect depression by analyzing the sentiments and opinions of users. In this review we focus only on the detection of mental state of users such depression using social media. In this paper we want to understand how individuals. We consider Depression because in todays world Depression affects most of the people over Worldwide leading to Suicide. Since Social media are somewhat latest aspect that power the relation among their use and feelings of aloneness and depression has not yet been accordingly researched. Most of the analysis on problem dated published in the last few years.

2. DEPRESSED RATE IN INDIA AND OVER THE WORLD: ACCORDING TO WHO

   WHO reports that approx.5 crore people suffer from Depression. The WHO report estimates that about 322 million people are suffered from depression over worldwide and nearly half of the populations are lived in South East Asian and Western Pacific Region. The total number of people that are living with depression are estimated increase by 18.4% between 2005-2015.

   WHO report in September 2012suggests that 75% suicides are committed in low and middle income countries. Lancet report in 2012 reports that India has the highest suicide rates in youth, age between 15-29. National Crime Records Bureau reports in 2013, 2471 students commit suicide because of failure in examination. The rate is high due to the lack of happiness and also patience in the individuals.

   It is noted that students with happy families has no depression on. Seeing that failing in exams and not able t o

   cope up with academics is the primary reason n for suicide. The Depression is a state that should be cured as soon as possible, because today in India Suicides rates are increasing day by day.

   ;

   Figure 1: State-wise Rates of Suicides.(Lanc et, 2015).

   The leading cause for suicide are, the highest rate family problems, illness, drug addiction etc.Lancet report, in 20 12 reports that A Student Commit Suicide Every Hour in Indi a. WHO in h is report Depression and other common Mental Disorders-Global Heal th Estimates mentioned t hat overall 7,88,000 people died because of suicide, and he number of suicides of students is approx. 8934 in 2015. In the leading five years 39775 students killed themselves and many of the suicides are unreported.

   Figure 12 : (National Crime Record Bureau, 2013)

   Figure 3: Different types of causes of Suicides. (Lancet Report,2015).

   From the above graph, he Lancet report show the percentage share of various causes of suicides. And by studying the above graph it seems that maximum number of suicides occurred due to family problems.

   A. Predicted number of Social Media use s from 2010-2021

   World Wide (According to Statistic Portal)

   The stats show s the total of social media users around the world starting with 2010-2016 including projections upto 2 021. In 2019 its far predicted that one therell be about 2.77 billions social media users a ll over the World. Social Media

   is very popular in North America. It occupies the first rank a cording to Statistic Portal.

   Figure 4: The number of Suicide Attempts (National Crime records Bureau,2015)

3. RELATED WORK

   Many of the studies are do ne in the field of mental illness a d many of the methods are applied by the researchers to predict the mental state of users using social media. The rates of Depression among the individuals are increasing day by day, so it is necessary to predict mental state of users (i.e Depression) and suiciderates. Social media platform such as twitter allow users to broadcast their opinions,sentiments,news and information to other people.

   Min Hane Aung et al., mention two methods to access the behaviour related to mental illness, first one is measuring real-world behaviour through sensors and the second is measuring behaviour intercede through technology.[4].Munmun de Choudhury etal., tells in his paper that over the arriving two decades,depression is projected ultimate the prominent explaination for disability in profitable nations. Taking Social media as a reliable tool to predict depression.[5]For the prediction of mental state of a usr data is, collected from various social media sites that a user joined them.

   Shararth Chandra Guntuku et al., mention two methods for collecting the data, by recruiting the participants to do a Depression survey to share their Facebook and Twitter data, or by collecting the data from public online sources.[6]

   1. Prediction through Traditional Methods

      As above Shararth Chandra untuku et.al.,2017 mention in his study one of the methods that prediction can be done by conducting a Depression Survey by collecting the data from public online sources[6].We have noticed that most of the researchers like Seung W. Choi et al., predict the mental states by using questionnaires (like Depressive Symptom Inventory- Suicide Subscale(DSI-SS),Beck Depression Inventory(BDI), Center for Epidermologic studies(CES-D))[7]. Scott R. Braithwait et al., examined Depressive Symptom Inventory-suicide subscale, Interpersonal needs Questionnair(INQ) and Acquired Capability for suicide scale(ACSS) for their research[8]. Munnmun De Choudhury et al., uses CES-D Questionnair as a primary tool to predict the depression levels in the crowdworkers. The participants who are giving CES-D test answered on a Likert-type-4-point scale[9]. Liu Yi Lin et al., calculate the depression using a 4-item- Scale that is developed by Patient-outcomes Measurement Information System(PROMIS)[10]. Sharth Chandra Guntuku et al., tells in his study that online forums and websites in which users make discussions are also a source of data collection related to mental health.[6]Centers for Disease Control and Prevention(CDC) tells in his report that Depression is a common and serious illness, affecting 1 out of 10 women 18-44 years. Feeling of postpartum depression are more intense and last longer than those of baby blues , a term used to describe the worry,sadness and tiredness. About 1 in

      9 women experience symptoms of postpartum depression.Postpartrum depression is also a major disorder that mostly women suffer after child birth.Munmun De Choudhury et al.,2014 examined the Facebbok to predict the postpartrum depression using Patient Health Questionnair (PHQ-9). This questionnair takes the responses from the last two weeks, based on experiences[9]. Postpartum Depression is the most important mental illness that a women faces after birth. Munnmun De Choudhury et.al., mention in his paper that according to CDC it is estimated about 12-20% of recent mothers faces a mood disorder). Various machine learning algorithms are used for the prediction and prevention of depression, so at what extent these algorithms are benificial against the suicide prevention.To do that Scott R Braithwaite et al., adopted various questionnairs,

      Capability for Suicide Scale(ACSS) [8].By adopting the questionnair pattern Psychiatrist fails to receive the whole information of the depressed patient. Social media overcome this limitation.

   2. Prediction through Social Media

      Munmun De Choudhury et al., opts the croud sourcing technique to extract the data from twitter and build a SVM classifier to predict the accuracy of depresssion [5]. Keumhee Kang et al.,uses crawling technique to collect the data form twitter and saves it into their database using open API. And crawling is done using keywords or real time streaming [11].Maryam Mohammed Aldarwish and Hafiz Farooq Ahmed developed a Web Application that characterize the social media users into one out of four Depression levels(Minimal, Mild, Moderate and severe).They colect the data from Facebook and Twitter and use the BDI-II questionnaire and analyze the collected data using several Text analysis APIs [2]. From the collected data, focus on the words that are used in the users comments and status[2], and for that Elvis Saravia et.al. and Peter Burmap et.al.uses (Term Frequency-Inverse Document frequency) TF-IDF and Pattern of Life Features(PLF) to capture the repetative words that a patient used and predict the emotions and behaviour of the patients [1][12].

      Munmun De Choudhury etal., performs a learning based analysis by considering forms and structures of sentences including words related to the human moods, and uses Support Vector Machine (SVM) to understand the relationship among them [5]. Min Hang Aung et al., in his study access the behaviour of human using mobile sensing, Supervised Learning method is used [4]. Munmun De Choudhury et.al. used Six different performance metrices(accuracy, precision and recall,F1, specificity and area under curve AUC) to calculate the performance of the SVM classifier. Linguistic Inquiry Word Count(LIWC) are used to categorise into positive and

      negative affects.LIWC is a tool that analysis the text that extracts information to calculates the depression among users based on the words in the dictionary[9]. Scott R Braithwaite et al.,done a analysis in python that predicts the mental state using scikit-learn library [8]. Glen Coppersmith et.al. uses two Language Models, first is the Traditional 1-gram LM (ULM) that calculates the each word as a whole, and second is the Character 5-Gram LM (CLM) that derives the sequences upto 5 characters [13]. Maryam Mohammed Aldarwish and Hafiz Faroq Ahmed build a depression model using RapidMiner and test the two classifiers i.e., SVM and Naïve Bayes classifier, and their system has the best precision and minimal accuracy and recall. [2] Braithwaite et al., build a model using Decision tree learning beacause it can provides a model with accuracy for many applications, and they used the leave-one-out cross- validation(loo-cv) to estimate the accuracy of the decision tree [8]. Pete Burmap et al.,built a number of baseline classifiers by using features that is extracted from twitter, and by using the Weka Learning Libraries they conduct the baseline experiments. They they build an ensemble classifier with the help of Rotation Forest Algorithm [12]. Quan Hu et al., predicts the depression in users using the Sina Weibo data and it uses the Chinese text analysis software Wen

      Depressive symptom Inventory-Suicide sub scale(DSI-SS),

      InVteorlpuemrseo7n,aIlssNueee0d1s Questionnaire(INQ),Acquired

      Published by, www.ijert.org 3

      Xinfor processing the text. For Features selection they applied Greedy Stepwise algorithm and build a classification model using Logistic Regression method and conclude that it is practical to predict a whether a users depressed or not by means of Social media [14].

      Yoshihiko et al., developed smartphone application

      Utsureko to collect the data from users and use the power of Deep Learning to build a depression detection model, and the results shows that their framework are able to predict the severe depression using individual histories gives high accuracy [15].In Social media people are connected to each other. It makes a bond. Eric Gilbert et al., predicts the Tie Strength with social media by using Firefox extension Grease monkey to collect the data and use LIWC to do analysis. SMDI(Social Media Depression Index are used and they conclude that SMDI can intently depict Centers for Disease Control and Prevention(CDC) explained stats on Depression.[16]. Today the personality of a user is most important. Personality poses the exact expression infront of others. Users personality can be predicted by their face book status, comments, images, pages they liked ,by analyzing their behaviour etc.

      Heather Cleland Wood et.al. mentioned in his study that using Nighttime social media and emotion investment can affects the sleep quality and levels of Depression Since by using social media at night time may adversely affects sleep in adolescence.[17] Heather Cleland Wood et al., findings conclude that the use of social media specially at bedtime is a major factor that affects adolescence sleep quality, and levels of anxiety, ie. Poorer sleep quality and increased anxiety and depression [17]. Kimberly A. Van Orden etal., study recommended that women might probabily experience many risks that shows the presence of thwarted belongingnessandperceivedburdnsomness [18].

   3. Mental Health Detection using Deep Learning

      Since Deep Learning is a method, with its beginning in Artificial nural Networks, is emerging in recent times as an imperessive tool for Machine Learning, encouraging to customize the field of Artificial Intelligence. George Gkotsis et al., National Language Processing of Electronic Health Recods is more and more getting used to learn about insnity and jeopardize behaviours in so much closer analyze than previously. In the study Author addressed the diffeculty of characterizing and commnly classifying user generated data on social network site Reddit for the prediction of mental health contitions. They manually researched data generated from a number of subreddits to gather the posts within the underline Theme. The obtained grouping of posts into Themes was furthermore evaluated by applying Topic Detection Algorithms, and results indicates that the Theme based grouping is suthentic. Then they applied two classification studies, a Binary classification to determine whether or not post contains mental health relevant data, and multiclass classification to discover the mental health condition, and they conclude that by applying Convolutionary Neural Network (CNN) in binary classification, they attain a accuracy of 91.08% [19].Richard Socher et al., in his study they introduced a Recursive

      Neural Tensor Networks and Standford Sentiment Treebank. And the combination produces a system for sentiment detection of a single sentence, which pushes state of the art by 5.4% for sentence classification. They compare their model to a number of models such as Recursive Neural Network, matrix-vector RNN etc.and they noticed that RNTN acquired a highest accuracy of 80.7%.[20]. Huijie et al., in his proposed work thay used the users micro-blog data and present a user-level mental stress detection and to do that they collect the ground truth data, and from the users tweets text and images they present a set of low level content attributes, statistical attributes, and a convolutional Neural Network is designed with cross auto encoders to accumulate these attributes and generate a user- scope attributes and they apply a deep neural network model to understand the higher level of attributes and predict mental stress. They test their model with four different data sets and their results shows that the model efficiently detect the mental stress of users by using miro-blog data. [21]. For message-level Twitter sentiment classification. Duyu Tang et al.,developed a Deep Learning system:Coooolll. Coooolll composed of two parts first is sentiment specific word embeddings(SSWE) and second is the state-of-art- handcrafted features for feature representation. The efficiency of the system Coooolll is verified. Coooolll yields a position 2 on Twitter 2014 test set among 45 systems of SemEval 2014 Task 9.[22].

      Table 1: Some posts that indicate Depession.

      Depression Indicative posts

      Hey bro, what happen , is everithing fine or not ? No.Nothing is fine, I dont understand why only me..why people always cheat only me

      Everyone taunts only on me.Why?….I m fedup of this.

      I wish I have someone that cares for me, understands me, love me Im sad.

      I loss my interest in all activities..I dont like anything..I want to be alone..

      Why I m not a first priority for a people.this makes me cry

      How is that poissible..how I can fail..I dont know what to do..there is no reason for me to alive.

      I diagnosed with Depression last week.

      I m fed up of all this, why cant people leave me alone, why they interfere all the time.

      In the above table we mentioned some of the coments that indicates Depression, and that are used by the people when they are not mentally well, and that and taken by the researcherchers to predict the users mental state. For Example: Hey bro, what happen , is everithing fine or not

      ? No.Nothing is fine, I dont understand why only

      me..why people always cheat only meThis coment indicates that a person is very disturbed and are close to Depression. How is that poissible..how I can fail..i dont

      know what to do..there is no reason for me to alive.this coment indicates the Suidality.

      Table 2: Comparitive analysis of most commonly used approaches used for precticting depression

      Author	Topic	Aim	Methods	Results
      Munmun De	Social Media as a		SVM classifier, Center for	SMDI can nearly reflect CDC characterised
      Choudhury et	Measurement tool of		Epidemiologic studies Depression	insights on depression
      al.,(2013)	Depression in	Measuring Depression	scale(CES-D), social media
      	populations		depression index(SMDI), Principal
      			Component Analysis(PCA),
      	Recursive Deep Models		Recursive Neural Tensor Network
      	for semantic		(RNTN), Matrix-Vector- RNN
      Richard Socher et	compositionality over a	Sentiment Prediction	(MV-RNN()	RNTN obtains 80.7% accuracy
      al.,(2013)	Sentiment Treebank
      Munmun De Choudhury et al.,(2014)	Characterizing and Predicting Postpartum Depression from Shared Facebook Data	Detect, Characterize and Predict Postpartum Depression	Patient Health Questionnair (PHQ- 9), LIWC	Postpartum Depression was best predicted.
      			Term Frequency inverse document
      	Machine Classification		frequency (TF-IDF), Linguistic
      Pete Burmapet	and Analysis of Suicide-	Identification and	Inquiry and Word Count(LIWC),	Results achieved an F-Measure of 0.728
      al.,(2015)	Related	classification related to	Principal Component Analysis,	overall.
      	Communication on	suicide	Support Vector Machines(SVM),
      	Twitter		Rule Based, Naïve Byes, J48	0.69 for suicidal ideation class.
      			Decision Tree, Rotation Forest
      			Depressive Symptom Inventory-
      Scott R.	Validating Machine		Suicide Subscale(DSI-SS),	Their results shows that it can be easily
      Braithwaite	Learning Algorithms for	To validate the machine	Interpersonal Needs	characterised the people who are at suicide risk
      et.al.,(2016)	Twitter Data Against	learning algorihms to	Questionnair(INQ), and Acquired	and who are not with the help of Machine
      	Established Measures of	predict suicide risks.	Capability for suicide	Learning Algorithms.
      	Suicidality		Scale(ACSS), LIWC: updated 2015
      			version, Scikit-learn library,
      			Decision tree earning
      Elvis Saravia et.al.,(2016)	MIDAS: Mental illness detection and analysis via social media	Predicting Depression	Center for Epidermologic studies depression scale, TF-IDF,PLF, Sentiment 140API, Random Forest Clasiifier: a main learning model	They built an online system that extracts the features of a user by concidering two mental disorders, and their system gives the minimal results,that can be used in future to predict the user behaviour more efficiently.
      	Identifying Depressive	Extracts Tweets from	SVM based Learning, Built a	The results shows that a multimodel that is
      Keumhee	users in Twitter using	Twitter that indicate	Lexicon by using Visual Sentiment	developed has high accuracy as compared to
      et.al.,(2016)	Multiodal analysis	Depression..	Ontoloy and Sentistrength	the existing methods, and can efficiently
      			dictionaries, LIWC, K-means	predicts the users mood.
      			Clustering Latent Fusion

      Maryam	Predicting Depression	Classification of users	BDI-II Questionnaire, create a	The performance of the model is calculated and
      Mohammed	Levels using Social	according to Mental	depression model using	they got the best precision and minimal
      Aldarwish	Media Posts	Illness.	Rapid Miner, SVM and Naïve	accuracy and recall
      et.al.,(2017)			Bayes classifiers
      	Multitask Learning for		Multitask Learning
      Adrian Benton	Mental Health		approach(MTL), Logistic	Results shows that the proposed model
      et,al.,(2017)	Conditions with Limited	Predicting Depression	Regression, Feed forward multilayer	performs better compared to LR models.
      	Social Media data.		perceptron Single task
      			Learning(STL),
      	Deep Mood: Forecasting		Long short term memory recurrent	The results shows that the developed framework
      	Depressed Mood Based	Predicting Depressed	neural networks, Utrsureko: a	to predict the severe depression based on the
      Yoshihiko Suhara	on	moods.	smartphone application, Ecological	individual histories is efficient and gives high
      et.al.,(2017)	Self-Reported Histories		momentary assesment (EMA)	accuracy.
      	via Recurrent Neural		Approach
      	Networks

      Table 2.summarises the different approaches used by the researchers to predict the Mental state of users, most of the approches are based on the uses of basic questionnaires(CES-D, PHQ-9, INQ, DSI- SS, BDI) to get the input from the users, then different machine learning algorithms were used to analyses the mental states of the users based on their input information. Later on some researchers have done the analysis of the mental states based on the input given by the users through mobile apps. These approaches are more efficient compare to the earlier approaches as they collect the user input during various phases of the day.

4. EXISTING SYSTEM

   It investigate the association of sleep quality and suicide attempt of Internet addicts. On the other hand, recent research in Psychology and Sociology reports a number of mental factors related to social network mental disorders. An NLP-based approach to collect and extract linguistic and content-based features from online social media to identify Borderline Personality Disorder and Bipolar Disorder patients. It extract the topical and linguistic features from online social media for depression patients to analyze their patterns.

   The analyze emotion and linguistic styles of social media data for Major Depressive Disorder (MDD). However, most previous research focuses on individual behaviors and their generated textual contents but do not carefully examine the structure of social networks and potential Psychological features.

   DISADVANTAGES

   • Although previous work in Psychology has identified several crucial mental factors related to SNMDs, they are mostly examined as standard diagnostic criteria in survey questionnaires.

   • To automatically detect potential SNMD cases of OSN users, extracting these factors to assess users online mental states is very challenging. For example, the extent of loneliness and the effect of disinhibition of OSN users are not easily observable

   • The developed schemes are not designed to handle the

   • The SNMD data from different OSNs may be incomplete due to the heterogeneity.

5. PROPOSED SYSTEM

   We argue that mining the social network data of individuals as a complementary alternative to the conventional psychological approaches provides an excellent opportunity to actively identify those cases at an early stage.

   In this paper, we develop a machine learning framework for detecting SNMDs, which we call Social Network Mental Disorder Detection (SNMDD). We propose an SNMD-based Tensor Model (STM) to deal with this multi-source learning problem in SNMDD.

   We propose an innovative approach, new to the current practice of SNMD detection, by mining data logs of OSN users as an early detection system. We develop a machine learning framework to detect SNMDs, called Social Network Mental Disorder Detection (SNMDD).We also design and analyze many important features for identifying SNMDs from OSNs, such as disinhibition, Para sociality, self- disclosure, etc. The proposed framework can be deployed to provide an early alert for potential patients.

   ADVANTAGES

   • The novel STM incorporates the SNMD characteristics into the tensor model according to Tucker decomposition; and

   sparse data from multiple OSNs.

   • The tensor factorization captures the structure, latent factors, and correlation of features to derive a full portrait of user behavior.

   • We further exploit CANDECOMP/PARAFAC (CP) decomposition based STM and design a stochastic gradient descent algorithm, i.e., STM-CP-SGD, to address the efficiency and solution uniqueness issues in traditional Tucker decomposition.

   • The convergence rate is significantly improved by the proposed second-order stochastic gradient descent algorithm, namely, STM-CP-2SGD.

   • To further reduce the computation time, we design an approximation scheme of the second-order derivative, i.e., Hessian matrix, and provide a theoretical analysis.

6. ALGORITHM

   To explore data mining techniques to detect three types of SNMDs

   1. Cyber-Relationship (CR) Addiction, which includes the addiction to social networking, checking and messaging to the point where social relationships to virtual and online friends become more important than real-life ones with friends and families.

   2. Net Compulsion (NC), which includes compulsive online social gaming or gambling, often resulting in financial and job-elated problems.

   3. Information Overload (IO), which includes addictive surfing of user status and news feeds, leading to lower work productivity and fewer social interactions with families and friends offline.

      We present a Stochastic Gradient-Descent Algorithm for CP decomposition of the SNMD-based Tensor Model, namely, SGD-CP-STM.To iteratively improve each element in the matrices according to the corresponding gradient. We present a stochastic gradient-descent algorithm for CP decomposition of the SNMD-based Tensor Model

      ,namely, SGD-CP-STM, to iteratively improve each element in the matrices according to the corresponding gradient. Specifically, let T ( ,V,W) be a matrix obtained from T by contracting V and, i.e.,

      where T ( ,V,W) RN×R (the same as U). The followinglemma first derives the gradient of each iteration.Lemma 1. The gradient of L with regard to U, i.e.,UL(T ,U,V,W), is equal toT ( ,V,W) + U((V,W) +

      2IR) + 1LaU, where (V,W) is the Hadamard product of V_V and W_W,i.e., (V,W)ij = (V_V)ij (W_W)ij, and IR is the identitymatrix of size R.Proof. The objective function L(T ,U,V,W) is comprised of three terms, and the derivative of 2 2_U_2 with regard toU is 2IR. For the first term, the CP gradient can be solvedby the following equation according to

      RTICCT – 2019 Conference Proceedings

      For the second term, i.e., 12 tr(U_LaU), the gradient for U

      is

      If the weighted adjacency matrix A is symmetric, Equation can be further simplified to 1LaU, and UL(T ,U,V,W) is equal to

      Therefore, the stochastic gradient descent algorithm updates U

      at the t-th iteration as follows.

      U(t) = U(t1) (t)(T (t1)(,V(t1),W(t1))

      +U(t1)((V(t1),W(t1)) + 2IR) + 1LaU(t1))

      Based on equation the gradient for V and W can be derived in the similar way as follows:

      VL(T ,U,V,W) = T (U, ,W) + V(U,W)

      WL(T ,U,V,W) = T (U,V, ) +W(U,V).

      Note that V(t) andW(t) are also updated similarly in each iteration.

7. DATAFLOW DAIGRAM

8. MODULES

   • Data Collection

   • Detect location of users post

   • Detect time of users post

   • Detect mental disorder users

   Suggestion to block user

   1. DATA COLLECTION

      Data collection is the systematic approach to gathering and measuring information from a variety of sources to get a complete and accurate picture of an area of interest. Data collection enables a person or organization to answer relevant questions, evaluate outcomes and make predictions about future probabilities and trends. you'll learn the many ways to import data into Python:

      1. from flat files such as .txts and .csvs;

      2. from files native to other software such as Excel spreadsheets, Stata, SAS and MATLAB files;

      3. from relational databases such as SQLite & PostgreSQL. This course teaches you to fetch and process data from services on the Internet. It covers Python list comprehensions and provides opportunities to practice extracting from and processing deeply nested data.

   2. DETECT LOCATION OF USERS POST

      Look for the gray location tag listed at the end of the post. Facebook posters can let their friends and followers know their location by posting a status update and selecting the icon to "add a location to post." The GPS or wireless connection attached to the device where the post was made determines city and state, or sometimes the exact venue. Get a history of the cities, countries, and other places a user has visited, as long as they took a picture there. Is possible anywhere to obtain this huge database of states/cities/regions? Would be interesting to have something similar in our app, but I don't know, where those lists get.

   3. DETECT TIME OF USERS POST

      Social media is one of the best ways to amplify your brand and the great content youre creating. But it isnt enough to just post content to social whenever you feel like it. Some times are better than others.

      So, what are the best hours to post on each social media channel?

      Unfortunately, there's no perfect answer. People browse each social network differently, and businesses may find different days and times work best for them. For example, while Twitter sees tweets perform well at hours like 3 p.m., Instagram sees certain posts perform well as late at 2 a.m. You worked really hard on those social media posts for an upcoming campaign, but do you have any idea how many people will engage? Learning the best times to post on social media is more than just turning a few clicks into a dozen. This data lets you understand your audience inside and out.

   4. DETECT MENTAL DISORDER USER

      It's widely recognized that psychiatric conditions like depression and anxiety disorders are based in the brain. Scientists have even started to discover which brain areas are involved in different conditions. For example, post-traumatic stress disorder (PTSD) seems to involve excessive activity in the amygdala, which is involved in processing fear, as well as low activity in certain parts of the frontal lobes.

      Much of the evidence for the role of specific brain areas in psychiatry comes from "brain imaging," which involves various ways of looking at the brain. Some technologies like PET imaging and functional MRI can measure the activity of the brain either at rest or while a person does certain tasks. Other technologies, like traditional

   5. SUGGESTION TO BLOCK USER

   We give a suggestion to block the users ID in their current social media. The blocking is based on reducing their time of usage in social media. We reduce the time of usage to 3hrs-6hrs according to their usage patterns, these are all done through the feedback area of the particular social media.

9. CONCLUSION

   The Lifestyle of people nowadays leads depression even in young generation. As the study indicates that the depression rate is increasing day by day. The reports of WHO and other organizations reflects that the consequences are worse, even lead to suicide. Our aim is to study the techniques used by the researchers so far to predict Mental State of the public. The study shows that most of the work is based on the input collected from the clients using questionnaires. It is observed that the usage pattern of Social Media users including their time of usage, their posts, and other social activities reflects the mood of the user which can be very helpful to analyses the Mental State of the users and predicting depression.

10. REFERENCES:

1. Elvis Saravia, Chun Hao Chang, Renaud Jollet De Lorenzo, Yi- Shin Chen,MIDAS: Mental Illness Detection and Analysis via Social Media, International Conference on Advances in Social Networks Analysis and Mining,Vol, August 2016.

2. Maryam Mohammed Aldarwish, HafizFarooq Ahmed, Perdicting depression levels using social media posts, 13th International Symposium on Autonomous Decentralized Systems, 2017.

3. Priya Nambisan,Zhihui Luo,Akshat Kapoor,Social media,Big data, and public health informatics: Ruminating behavior of depression revealed through Twitter,48thHawaii International Conference on

   System Sciences, 2015.

4. Min Hane Aung, Mark Mathews, Tanzeem Choudhury, Sensing Behavioral Symptoms of Mental Health and delivering personalized interventions using mobile technologies, Wiley Periodicals, Vol 34, PP 603-609, 2017.

5. Munmun De Choudhury, Scott Counts, Eric Horvitz,

   Social Media as a Measurement Tool of Depression in Populations, WebSci, May 2013.

6. Sharath Chandra Guntuku, David B Yaden, Margaret L Kern, Lyle H Ungar1 and Johannes C Eichstaedt,

   Detecting depression and mental illness on social media: an integrative review, Current Opinion inBehavioral Sciences,Vol 18, PP 43-49,2017.

7. Seung W.Choi, Benjamin Schalet, Karon F. Cook, David Cella, Establishing a common metric for depressive symptoms: Linking the BDI-II,CES-D, and PHQ-9 to PROMIS depression, Psychological Assesment, Vol 26, PP 513-527, 2014.

   MRI, measure the brain's structureits size and shape.

8. Scott R Braithwaite, Christophe Giraud-Carrier, Josh Machine Learning Algorithms for Twitter Data AgainstEstablished Measures of Suicidality,JMIR mental health, Vol 3, 2016.

9. Munmun De Choudhury, Scott Counts, Eric J. Horvitz, Aaron Hoff, Characterizing and predicting postpartum depression from shared Facebook data, Social Technologies and Well-Being, Feb 2014.

10. Liu Yi Lin,B.A.,et al., Association between social media use and depression among U.S young adults, Wiley Periodicals, Vol 33, PP 323-331, 2017.

11. Keumhee Kang, Chanhee-Yoon, Eun Yi Kim(2016).

    Identifying Depressive Users in Twitter Using Multimodal Analysis, BigComp, 2016.

12. Peter Burmap, Gualtiero Colombo, Jonathan Scourfield, Machine Classification and Analysis of Suicide-Related Communication on Twitter, 2015.

13. Glen Coppersmith, Mark Dredze, Craig Harman,

    Quantifying Mental Health Signals in Twitter,Workshop on Computational Linguistics andClinical Psychology, PP 51-60, June 2014.

14. Quan Hu, Ang Li, Fei Heng, Jianpeng Li, Tingshao Zhu, Predicting depression of social media user on different observation windows, International Conference of Web Intelligence and Intelligent Agent Technology, 2015.

15. Yoshihiko Suhara, Yinzhan Xu, Alex Sandy Pentland, DeepMood: Forcasting depressed mood based on self-reported histories via Recurrent Neural Networks, International World Wide Web Committee, April 2017.

16. Eric Gilbert and Karrie Karahalios, Predicting Tie Strength With Social Media, CHI 2009, 2009.

17. Heather Cleland Woods, Holly Scott, #Sleepyteens: Social media use in adolescence is associated with poor sleep quality, anxiety, depression and low self- esteem,Journal of Adolescence, Vol 51, PP 41-49, 2016.

18. Kimberly A. Van Orden, Tracy K Witte, Kelly C Cukrowicz, Scott R. Braithwait, Edward A. Selby, Thomas E joiner,The Interpersonal Theory of Suicide,American Psychological Association,Vol 117,PP 575-600. 2010.

19. George Gkotsis, Anika Oellrich, Sumithra Velupillai, Maria Liakata, Tim J. P. Hubbard, Richard J. B. Dobson& Rina DuttaCharacterization of mental health conditions in social media using informed Deep Learning,2017.

20. Richard Socher, Alex Perelygin, Jean Y. Wu, Jason Chuang, Christopher D. Manning, Andrew Y. Ng andChristopher Potts, Recursive Deep Models for Semantic Compositionality Over a SentimentTreebank,Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, PP16311642, 2013.

21. Huijie Lin1,2 , Jia Jia1,2, Quan Guo3 , Yuanyuan Xue1 , Qi Li1 , Jie Huang1 ,Lianhong Cai1,2, Ling Feng,User-Level Psychological Stress Detection from Social Media Using Deep Neural Network, MM14, 2014.

22. Duyu Tang, Furu Wei, Bing Qin, Ting Liu, Ming Zhou, Coooolll: A Deep Learning System for Twitter Sentiment Classification, Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), PP 208212, 2014.

23. Adrian Benton, Margaret Mitchell, Dirk Hovy, Multitask Learning for mental health condition with limited social media, Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics, Vol 1, PP 152-162, April 2017.

24. Blei, D. M., Ng, A. Y., and Jordan, M. I. Latent dirichlet allocation, The Journal of Machine Learning ,Research 3, PP 9931022, Jan 2003.

25. NikolinaBanjanin, Nikola Banjanin, Evan Dimitrijevic Igor Pantic, Relationship between Internet use and Depression: Focus on physiological mood Oscillations,social networking and online addictive behavior, Computers in Human Behavior, Vol 43,PP 308-312, 2015.

26. PK Singh, MS Husain, Methodological study of opinion mining and sentiment analysis techniques, International Journal of Soft Computing, 5(1), 11, 2014.

27. MS Hussain,Analylitical study of feature extraction techniques in opinion mining, Computer Science, 2013.

[28]N Khan, MS Husain, M R Beg, Big Data Classification using Evolutionary Techniques: A Survey, IEEE International Conference on Engineering and Technology(ICETECH), PP 243-247,2015.

[29]T S, H M Shahid.Opinion mining with Spam Detection using Real Time data on cloud, International Journal of Advanced Research in Computer Science 8(7), PP 910-914, 2017.