Review on Machine Learning Techniques to predict Bipolar Disorder

: Bipolar disorder, a complex disorder in brain has affected many millions of people around the world. This brain disorder is identified by the occurrence of the oscillations of the patient’s changing mood. The mood swing between two states i.e. depression and mania. This is a result of different psychological and physical features. A set of psycholinguistic features like behavioral changes, mood swings and mental illness are observed to provide feedback on health and wellness. The study is an objective measure of identifying the stress level of human brain that could improve the harmful effects associated with it considerably. In the paper, we present the study prediction of symptoms and behavior of a commonly known mental health illness, bipolar disorder using Machine Learning Techniques. Therefore, we extracted data from articles and research papers were studied and analyzed by using statistical analysis tools and machine learning (ML) techniques. Data is visualized to extract and communicate meaningful information from complex datasets on predicting and optimizing various day to day analyses. The study also includes the various research papers having machine Learning algorithms and different classifiers like Decision Trees, Random Forest, Support Vector Machine, Naïve Bayes, Logistic Regression and K-Nearest Neighbor are studied and analyzed for identifying the mental state in a target group. The purpose of the paper is mainly to explore the challenges, adequacy and limitations in detecting the mental health condition using Machine Learning Techniques.


Introduction
World Health Organization WHO implies, a person having healthy mind and physical fitness is a healthy person. Any changes in thought process and mental health are some of the age related process and changes across the world. Depression and anxiety are mental health disorder associated with unhealthy mind. As the age increases, the consequences and vulnerability associated with depression and anxiety also increases [3]. The development and advances in big data Analytics and Technology results in more attention in prediction of disease. Various studies and researches on large number of dataset has been conducted automatically to improve the accuracy of risk classification instead of selected characteristics previously [1]. Patients having bipolar disorder significantly experiences day-today and week-to-week swings in mood. This instability in mood increases the relapse of disease and reoccurrence of risk with time, this indicates that the disease is still active. The purpose of monitoring and symptom prediction of disorder is to investigate and correlate the symptoms of disorder [8].
Machine Learning Techniques are increasingly present in almost all systems that Process and gather bulk amounts of data. The field of medicine is largely benefited by Machine Learning. Machine Learning algorithms design the regression and classification models that help in different disease diagnosis, drug recommendation, drug administration and so on. ML is the process of creating certain models and algorithms to predict values based on different features [13]. This review paper analysis ML technique for bipolar disorder and clinical procedures. The Data of healthy and unhealthy person are reviewed from a certain survey to apply prediction algorithm.

Categories of Bipolar Disorder
Bipolar disorder is a lifelong mental illness that occurs due to episodes of Mania and depression. Even after getting treatment for the illness people continues to have symptoms for it. Each type of disorder is identified and treated differently depending upon the type of disorder as shown in

1). Planning and Background Analysis:
In this phase all the information regarding the individual and his illness is gathered which includes various qualitative factors like personal information, demographic information, sociodemographic characteristics, their Symptoms, and disease. Age, gender, their past history, chronic medical conditions, family status and environment, marital status, job security etc. are analyzed for detecting mental condition like depression and anxiety in older people [3]. These attributes are used as predictors in automated system for disease prediction [3].

(3.3). Data Preprocessing-Feature Selection:
The raw data collected in data collection phase is preprocessed in an understandable format by two methods, namely data cleaning and data transformation. This is used to differentiate the various behaviors of the patients and to select the feature on the basis of which medical assistance is given. For describing and demonstrating depressive and non-depressive comments and posts, different features were extracted from user's post having psycholinguistic features.

.1) Random Forest
This classifier is a supervised classification method also known as ensemble classification method. The result of individual prediction is averaged out using predictor attributes. They are trained by bagging method that consists of random sample subsets. By sampling the training data, the method fits a decision for each tree and aggregate the result. Many base classifiers are provided by Random Forest. Values are inputted randomly in each tree and their distribution is equal in each tree. [20].

(3.4.2) Support Vector Machines (SVM)
This classifier is a linear and non-probabilistic binary classifier used to classify data for anomaly detection. Among other ML Techniques this modal is very popular as well as costly to compute [21]. SVM is used for reducing noisy data for good results and is able to make decisions [22]. This classifier is a Statistical model used for regression challenges and classification. SVM is a supervised Machine Learning Algorithm in which for n number of parameters, each n parameters are plotted in n-dimensional space and assigned a specific coordinate. The classifier built a hyper plane into high dimensional feature space to isolates data in two classes. This finds a hyper plane for the purpose to separate close training datasets. SVM creates hyper plane by calculating the best possible margin (which is the difference between hyper plane and support vectors). Support vector divides dataset in higher n dimensional space.  approach. This makes the hierarchical tree from the training dataset. The states of decision tree are to divide the hierarchy of data having different characteristics. For example, in text documents classification, roots are mainly identified in terms and internal individual nodes are subdivided into its children in view of the yes or no of a term in Ensemble. Ensemble methods use multiple learning algorithms of decision tree for better predictive performance.

Investigating ML techniques for the prediction of Bipolar disorder
The review aims to study and give a clear and concise literature investigating Machine Learning ML techniques for the prediction of Bipolar disorder. The literature review aims to reduce the occurrence and prevalence of the anxiety disorders by effective early prediction. This results in significant minimization of hospitalization, improving their quality of life and reduces their health care bills to a large extent. The Literature review has three stages as shown in    [10] in this a AD novel prediction model is proposed for prediction of anxious depression in real time tweets. This studies the mixed disorder od anxiety and depression associated with thought process, lack of sleep and restlessness. Ezekiel Victor et al. [11] suggested a methodology which evaluates ML techniques for detecting depression that require minimal human intervention.in terms of data collection and data labeling. Emmanuel G et al. [12], Conducted a comparative literature search using ML Techniques to predict specific types of stress and anxiety disorder and to develop certain tools that assists doctors in prediction of mental health and support in caring patients. Md. Rafiqul Islam et al [13] implements ML techniques to identify and implement a quality solution for mental disorder problem by studying Social media users comments and posts specially Facebook users. For this they monitor their attributes, feeling and behavior. Their mood swing patterns while they communicate with other user in online communications. Adrian B. R. Shatte et al. [14] according to this paper there is scope of Machine Learning in the area of Psychology and Mental Health, and evidently focus on prediction and diagnosis of mental health condition. Liana C.L.et al. [15] Their study and findings gives an early neuroimaging techniques for clinical assessment in young adults irrespective of objective and qualitative estimation of Psychopathology. Alicia Martinez et al. [16] Abd Rehman et al. [22] This paper aims at providing a guideline for further research in the direction of health care prediction system using Machine Learning Techniques. The electronic dataset on health records Provides a valuable information about the health risk and its predictions. The ML applications and methods has provided benefits in treatment, support and diagnosis of research and clinical administration. A. Khatter et al. [23], how to deal with pandemic situations and save lives in Lockdown. Students adjusted them well at their homes with restrictions and with a that one day their life will again be normal as before. They cope up with online teaching Learning method, online exam patterns, admissions to higher studies and summer internships etc. MS. Purude Vaishali Narayanrao et al. [25], include different approaches to predict heart attack, peer pressure and depression. The data collection mechanism includes questionnaires and surveys with different people, their social media posts, text messages and verbal communication and facial expressions. Ela Gore et al. [26] aim of the paper is to present commonly used algorithms and methods to describe the performance that act as a guide in selection of appropriate model. The alternates and possibilities of ML helps to bridge the gap between psychiatrist and patients to reveal the embarrassment of patient in critical shortfalls. Norah Saleh Alghamdi [28] study the use and benefits of Artificial intelligent application which uses text analytical tool for mental health support. This app uses different technologies and innovative sensors built in smart devices. By using camera sensors and performing self testing scales it detects anxiety and depression. U Srinivas ulu Reddy et al. [29] by applying ML Techniques, a stress analysis pattern is studied in working employees and to narrow down the stress levels. For the study they used 2017 mental health survey that includes responses of technologies working employees. Vidhi Mody Pruthi Mody [30] facilitates a specialized care and emotional support for mental health people with the help of Machine Learning algorithms and Advanced Artificial Intelligence Techniques. Shahidul Islam Khan et al. [31] The study examines a classification algorithm that is used to predict mental health disorder.

Future Challenges in Mental Health Detection
The mental health disorder is difficult to categories because various feature selection processes are implemented by Researchers which is a major challenge in this study. The quality of dataset and its interpretation is a challenge. The data collected from the various devices should be very accurate and precise. Imprecise data from devices will lead to failure of the proposed system. The security, Privacy and ethical issues are important challenges in this field. To avoid issues related to privacy, safety precautions need to be taken, such as user authentication mechanisms and encryption of data. The information available in Online Social Networks [22] provides a huge or bulk of data having immense potential that is to be explored in modern research. For this we extract millions of data to understand the phenomenon selected for a study. The researchers focused on detection of mental health problem through several findings to be referred by researchers for future studies.
i) Initially few studies on mental health are found very informative like people with this disorder isolates themselves do not communicates with other peoples. They are quite simple and do not interact easily. Their social life is not normal as compared with non-stressed people. [28]. It's a challenges to make them feel that they are also normal humans.
ii) People with depression are involved in negative emotions and religious judgements [14]. They are involved in their own self. iii) The major challenge is of language barriers as they use different languages in mental health problem detection. During data analytics, in Online Social Networks it was diagnosed that people with depression behaves differently on various situations [22]. iv) Detection of Mental health involves several challenges in non-face-to-face communication and human computer interaction [22]. v) The use of machine learning(ML) can help them to understand and determine the possibility of an existing mental health state behind the words and languages in [14]. vi) The Privacy and security policies are the challenges faced by many researchers during their data preparation as due to the collection of public user data, such as those collecting from Twitter. Some of these Challenges are summarized as follows: - The Quality and interpretation of data set and modal.  Detection of mental health with time.  Multiple categories of mental health problems.  Preprocessing of data.  Data quantity and generalizability.  Data sparsity and ethical code.

Conclusion:
This review paper concludes that if the clinical heterogeneity of the samples of patient's data having bipolar disorder is given then by using machine learning techniques will provide researchers and clinicians with great insights in the fields such as diagnosis of diseases, their personalized treatment and prognosis orientation. Machine learning techniques for the prediction of stress and mental health condition will gives significant response and this can be studied and explored for further research objectives. Over the time, if we do not control the emotional conditions, anxiety will become worse day by day and turns into a pathological situations and that is quite challenging to treat. These mental health disorders result in harm to the human body as it leads to suppression of the immune system, which then increases the chance of susceptibility to various infectious diseases, increase in blood pressure, and diabetes.