Research Paper

Introduction

This paper explores human behavior to detect their depression levels in the digital world by analyzing their behavioral patterns. The digital world has reduced the interaction time between human encounters and made it an easily available interaction via social media. More than 80% of the human emotions are shared in social media since the physical human interaction has fallen drastically that everyone prefers social media rather than healthy human interactions. Social media is serving as a double edge sword when human emotions are considered since it has the ability to destruct happiness and also the ability to lift a person from depression and related issues. This leads to focus and emphasize on ethical usage of social media since 90% of the user lot are taking it for granted maintaining too many profiles on different pseudo names. As the platform has grown bigger many types of research have been taking place for understanding the current psychological situations of the world population.

Literature Review

The emotional analysis has gone into a research-level topic many researches have incorporated linguistic processing and content interpretation to understand the behaviors of an individual. This paper studies the human behaviors and their emotions by considering the sources like customer reviews on internet articles, social media postings, stock fluctuations, product reviews, newspaper reactions, etc. This paper uses K-mean clustering and Neural networks to efficiently understand the human behaviors which give the true values for false rejections and true rejections. Back Propagation Neural Networks helps in gathering the related patterns that define a particular emotion and thereby the subject emotions get narrowed down and if the subject is a close study material then it would be easy to diagnose the subject with the required solution

Deep Learning Neural Network in depression analysis

The recent advances in Data Science and analysis techniques lead to many groundbreaking researches and depression prediction is one such research where the technology is giving enough insights in the medical field. Digital world is the home ground for many depression related issues since the major population is spending too much of time on the social media and digital gadgets that they prefer sharing their dark sides and emotions in the form of social media postings and writings in wide range of platforms like blogs and community groups.

Link mining

Link mining is one of the prominent technologies in the data mining where the data instances are linked through wide range of data models that can categorize the content into various groups based on the prime focus and subject of the sentences. Unstructured data is not considered in this aspect since the Link mining is only able to link the words and sentence prime focus. There are also many individuals who focus on expressing their emotions via images and posts that contains unorganized data such as images and videos which are handled by models that filters unorganized data.

The link mining techniques that are widely used in finding the patterns related to depression and the sub topics around can be discovered by the following tasks

Link-based object ranking

Link-based object classification

Group based Algorithm

Entity resolution

· Link prediction

· Subgraph discovery

· Graph classifications and

· Generative models for graphs

The probabilistic relational models are used to get the links between the model objects and prime subject key words that can categorize a content words into the categories such as depression, fun, official etc. Each link mining tasks are applied in isolation to detect all the possible degrees of analysis in a particular task of the subject. Predictive analysis techniques are also applied in guessing weather a particular post of a user leads to future repetition of similar posts or its just an out of the moment type of posting. Likewise, various categories of the emotional levels are listed and validated to approach the base line value of the depression level.

The author XXX proposed a framework for link distribution models which is a model that supports discriminative models that links the distribution and related objects. The domains that are included in this include logistic classification over statistical values of the occurrences of repeated targeted word list. Link attributes and single flat classifier contents are linked together to gather the required information in solving the patterns that are usually witnessed in the social media postings and comments. Based on the links and objects the accuracy of classification is validated that reflects the effectiveness of the models that are run on particular data sets. The same data sets is run through all link based tasks that are well suited for a problem and the results are recorded to gain the valuable insights that helps in categorizing the mind set and extent of the depression level.

Document gathering is an important functionality in analyzing the behavior patterns, the strength of the hits and page ranks are registered to observe the basic behavior of the user. This record gives an analysis of the user’s behavior from one page to the other. The user web-based search and its related pages are observed to view the connecting link based pages that leads to different categories of the user behavior. Stable link investigation techniques are implemented in understanding the behaviors of the user which leads to automation analysis techniques. Since the stable link investigation helps in automation of the user behavioral patterns it is most important for the analyzing tools to effectively record the initial patterns that leads to automation process effectively.

Subspace HITS algorithm

This is an algorithm that is built by considering the motivation of observed subspaces spanned by eigenvectors a statistical methodology that is used to connect the matrix value of the problem set that is established. Experimental performance of various algorithms are considered to check the diverse nature of the results and the outliers and unrelated data is eliminated thus making effective scope for the analytics teams to come with good patterns that exactly brigs out the needed patterns in solving the problem. Thus the relation between Latent semantic indexing and HITS is established to verify the effectiveness of the algorithms outputs.

The entities of the relational domains are considered in enhancing the problem solutions that can give proper insights in expanding the research that gives scope for the automation of processes that needs to be repeated. Collective link classification is implemented to predict and classify the links corresponding to the given target metrics of the depression data set. This classification leads to establish an effective relation between the user patterns and the patterns recorded in the model. Classification accuracy is improved over the flat models by following two patterns

1. Similarity template and

2. Transitive template

Similarity template classifies the links and objects that share a common ground such as the links that have common end point which leads to subsequent web searches. This relation is called as the classification of template based on behavioral graph based pattern. The Transitive template considers triples of objects and forms various triangles that allow the tools in understanding the links between various related objects that form a group of triangles which serve in categorizing the patterns and thus the behavioral category is finalized.

Vicinity of Nodes

Inquiry is a common connection entity that connects various attributes of problem solving, it helps in creating techniques that explains the prediction of views in detecting the vicinity nodes of a network. Data related to future associations can be extracted by network topology and also discovering of node vicinity can be done which helps in recording the functionality of the Nodes. Training time plays an important role in handling the vicinity Nodes and hence the irregular indicators needs to be avoided to have effective training time. Quick calculations and nearness measures plays an important role in handling the effectiveness of the training time.

Weighted Network Model

Weighted network model is built to record the user actions on web specially social media platforms which considers the user associations which is stored in a special content called as information semantic content. This strategy utilizes the weighted system model that reflects the common grounds of user associations, system clustering, module detection that promotes learning graph. The topological properties such as user activity in the form of social media posting and commenting are analyzed to look for the semantic content that maps the available data in the already recorded behavioral patterns. These patterns are matched with the users actions to categorize the system models that fits the user network; all this weighted network interactions together form into a better assessment methodology that helps in analyzing the behavior.

Crowd sourcing philosophy is considered as one of the primary outliers on a major scale and considered as one of the needed measure that falls into the depression models since the instincts to follow a crowd depends on the existence of the trait in minor scales. A probabilistic model is constructed around this behavior which can demonstrate the depression levels. Primarily this is calculated on number of crowd based posts that attracted a particular user; more the number of similar posts yield to more probability of depression levels and vice versa. The signs that are used in the user behavior are social activity and language showed on the content. An online depression networking file is generated that classifies various levels of depressions. The classifier can easily estimates the degree of depression levels.

A two-stage analysis system is build to categorize positive and negative opinions of the user activity that can filter the traffic that is needed in the specific problem. Self-sorting methodology is applied that sorts the traffic based on the given parameters. All the posts are gathered, and word recurrence are calculated that can give an estimation depression level. Self sorting is the novel system in demonstrating the approach of the users based on their mental state. This helps the analyzers to distinguish the population between the influential users and the regular users.

Inter sentimental language pattern is one other method that helps in analyzing the user behavior on web to understand the state of the person. This pattern helps in connecting the relationship between the words and sentences in the given data, this gives more scope of the exact data in the form of words and sentences. Content mining structure is built by forming an association between traditional association techniques and sentence limits. Execution is easily predicted in a pile of data sets therefore the possibility of getting right results is more favored at all instances.

The study of connected data is more favoured than the un-connected data in many ways. Semantic closeness measure is a prime metrics in understanding the process of segmenting the right set of words that match the depression related data set. The target data sets are filtered to form a connection that establishes exact difference between the normal user and influential user. Light weight data structure and composition of data plays a major role in segregating the underlying differences between the depressed user and an average user.

With the use of predictive analysis, it is possible to combat social media depression. It gives a great advantage to the counselors to understand the patients and cure the disease in a more robust way. Here, this technique uses a simple formula by comparing the past data with the current data in order to predict future possibilities. For instance, by using machine learning algorithm, researchers identified some unique behaviors to identify depressed individuals. These early warning signs include mentions of loneliness or isolation, such as “alone”, “ugh” or “tears” as well as the timing and length of posts. Other tell-tale clues include an increase in the use of first-person pronouns – like “I” and “me” – which “suggest a preoccupation with the self” in public posts.

The result of the research was intriguing. They have identified some warning signs of depression in the participants from their social media posts over 3-month period. This is a great success and by using this technique they can identify the individuals with speed and accuracy.

Medical research states that social media addiction is close to that of the drug addiction with same set of mental neurons getting activated in both the cases. The more the digitalized tha life gets the more are the chances of getting depressed. The digital world has paved paths to many innovative tools and technologies but the inability to handle the right tools at right places have damaged the society in a higher degree than ever before. This kind of attitude towards the technology and digitalization can lead to adverse effects in the future generations.

The current generation kids are more into artificial stimulants like television and video games which are making them more anxiety and more addicted to the indoors. Children getting obese too with the kind of digital habits they are maintaining. They are also not even engaging their times in paly grounds and hence the digital and corporate heads have to take ethical responsibility in handling this problem. The video game makers play an important role in handling this situation since the current day video games are more of violence than entertaining games.

The graph below states the extent of video game manufacturers targeting the young innocent brains. As the movie industry giving certifications for the violent related certifications like A, the same should also be applied to the video games that encourage violence and killing. The recent video game that has got world popularity just because of the violence , PUBG which is based on killing and shooting. These kinds of games are also tempting for the gun violation in many advanced countries like USA and Canada. The regulations need to be made scrutinized in order for the manufacturers to focus on the content that can nurture the young brain in a constructive way.

References

Fatima, I., Din, B. U., Khan, S., Majed Al‐Saeed, Hafiz, F. A., & Mumtaz, R. (2019). Prediction of postpartum depression using machine learning techniques from social media text. Expert Systems, 36(4)

Singla, S. K., Kaur, R., & Bala, A. (2018). Prediction of depression in human behaviour using deep learning neural network. International Journal of Health and Economic Development, 4(2), 16-23. Retrieved from https://search.proquest.com/docview/2198408377?accountid=10378

Mumtaz, W., Syed Saad, A. A., Mohd Azhar, M. Y., & Aamir, S. M. (2018). A machine learning framework involving EEG-based functional connectivity to diagnose major depressive disorder (MDD). Medical and Biological Engineering and Computing, 56(2), 233-246. doi:http://dx.doi.org/10.1007/s11517-017-1685-z

Wang, W., Li, Y., Huang, Y., Liu, H., & Zhang, T. (2017). A method for identifying the mood states of social network users based on cyber psychometrics. Future Internet, 9(2), 22. doi:http://dx.doi.org/10.3390/fi9020022

Alharthi, R., Alharthi, R., Guthier, B., & Saddik, A. E. (2019). CASP: Context-aware stress prediction system. Multimedia Tools and Applications, 78(7), 9011-9031. doi:http://dx.doi.org/10.1007/s11042-017-5246-0

Farnadi, G., Sitaraman, G., Sushmita, S., Celli, F., Kosinski, M., Stillwell, D., . . . De Cock, M. (2016). Computational personality recognition in social media. User Modeling and User - Adapted Interaction, 26(2-3), 109-142. doi:http://dx.doi.org/10.1007/s11257-016-9171-0