1

ML AND HUMAN BEHAVIOR PREDICTION ON SOCIAL MEDIA

Understanding Human Behavior on Social Media through Machine Learning

A Literature-Guided Approach

Abid Hossain Rion and Mashrufa Meghla Any

Department of Artificial Intelligence & Behavioral Computing, MindBridge Innovative Network

May 10, 2025

 2
ML AND HUMAN BEHAVIOR PREDICTION ON SOCIAL MEDIA

Abstract

Social media platforms like Facebook and YouTube have become a daily interaction partner of

humans for opinion sharing, entertainment, and behavioral expression. With the vast increase of

human interaction data in social media, machine learning (ML) uses powerful tools and

techniques to predict human behavior based on their online activity. The purpose of this study is

to explore how machine learning predicts human behavior on social media by reviewing recent

literature. This study will highlight neural networks and support vector machines by examining

their effectiveness in social media contexts. This literature review study will analyze behavioral

prediction tasks, including sentiment analysis, mental health detection, and personality inference.

This paper provides a complete view of current trends, challenges, and moral considerations. The

findings suggest that while ML can reveal significant behavioral insights, data privacy, model

bias, and interpretability remain pressing concerns.

Keywords: machine learning, human behavior, social media analysis, behavior prediction,

mental health detection

 3
ML AND HUMAN BEHAVIOR PREDICTION ON SOCIAL MEDIA

Understanding Human Behavior on Social Media through Machine Learning

In this digital era, social media has changed the way individuals communicate, express

themselves, and share their opinions. There are so many social media platforms on the internet

such as Facebook, Twitter (X), and YouTube where people spend a vast amount of time in their

day-to-day activities. These social media platforms generate vast amounts of human behavioral

data every second which creates a digital footprint that reflects users’ thoughts, emotions, and

social patterns. This massive amount of user-generated data has drawn the attention of

researchers aiming to understand human behavior through data-driven methods. Among the tools

used to make sense of this data, machine learning (ML) stands out as one of the most powerful.

By analyzing social media activity, machine learning techniques can automatically detect human

thinking patterns and predict outcomes. These advanced machine learning techniques can

identify complex human sentiments and personality detection and analyze them to identify

mental health concerns. Increasingly, researchers are using advanced models like neural

networks and support vector machines (SVMs) to help understand and predict human behavior

on social media, and these tools have already shown some promising results. However, the

benefits of machine learning (ML) on social media are significant, but some challenges remain.

Issues such as data privacy, algorithmic bias, and lack of comprehensibility present ethical and

technical limitations. These circumstances raise consequential questions about how machine

learning should be used to study human behavior responsibly and accurately. This paper aims to

explore how machine learning is applied to predict human behavior on social media by

reviewing recent academic literature. By focusing on key machine learning (ML) methods and

their performance in behavioral prediction tasks, this study highlights both the potential and the

limitations of using algorithms to understand human nature in the digital world.

 4
ML AND HUMAN BEHAVIOR PREDICTION ON SOCIAL MEDIA

Literature Review

Phan et al. (2016) gave a novel model about deep learning, which is Social Restricted

Boltzmann Machine (SRBM+), developed for predicting and explaining human behavior in

health-focused social networks. In contrast to conventional approaches, SRBM+ combines self-

motivation, explicit, and implicit social influences, and environmental factors into its structure. It

offers not only accurate behavior prediction but also clarity by quantifying the impact of various

behavioral determinants. The study found that including both personal and community-based

historical data not only enhanced the accuracy of behavior predictions but also helped to build

user trust by providing clear and understandable explanations (Phan et al., 2016). Gulhane and

Sajana (2021) studied and analyzed human body parameters, food habits, and social influences

using various machine learning (ML) techniques. Their study highlighted the effectiveness of

Convolutional Neural Networks (CNNs) for automatic feature extraction from data sources like

EEG and ECG. The study discussed how combining physiological data with online behavior can

enhance disease prediction systems. For detecting the primary signs of conditions such as

diabetes, cancer, and cardiovascular disorders, they examined mixed approaches involving both

sensor data and social behavior (Gulhane & Sajana, 2021). The study also identified several

machine learning (ML) models that were used to predict human behavior based on mental and

perceptual data. The models are LASSO regression, elastic-net, and support vector regression.

The effectiveness of these models can change based on the sample size and feature complexity,

said to the authors. However, these models also offer good accuracy (Gulhane & Sajana, 2021).

Han et al. (2023) extracted linguistic and psychological features from Weibo posts and studied

personality prediction. Their machine learning (ML) model used a method called multi-objective

extra trees (MOET), which looks at the words humans use and then matches these words to the
 5
ML AND HUMAN BEHAVIOR PREDICTION ON SOCIAL MEDIA

Big Five personality features using some special psychology-based word lists. The study showed

that adding domain knowledge to the model improved both interpretability and accuracy. The

study found that including domain knowledge improved prediction, and features from the

Culture Value Dictionary were the most useful. It shows that a person’s culture plays an

important role in understanding their personality (Han et al., 2023). The authors also underscored

the need for explainable AI in psychological applications. Whereas many models provide

accurate predictions. However, the interpretability of the outcomes remains limited, which poses

challenges for their adoption in clinical or high-stakes settings (Han et al., 2023).
 6
ML AND HUMAN BEHAVIOR PREDICTION ON SOCIAL MEDIA

Methodology

This study employed a narrative literature review approach to explore how machine

learning (ML) techniques predict human behavior on social media platforms. Narrative Review

is a method where various research data are gathered together and analyzed, especially when the

topic is new and involves a combination of new topics, such as machine learning and human

behavior analysis, It does not involve statistical or experimental analysis but rather simply

explain the subject.

The literature was selected based on relevance to the study's core themes: machine

learning applications, social media behavior prediction, personality inference, sentiment analysis,

and mental health detection. The research was mainly searched using Google Scholar and

PubMed, by using different combinations of keywords like “machine learning,” “human

behavior,” “social media,” “personality prediction,” and “mental health detection.”

Additionally, specific peer-reviewed articles were included based on their contribution to

the research questions. These include: a deep learning model using SRBM+ for behavioral

prediction (Phan et al., 2016), a review on machine learning (ML) techniques for analyzing

social behavior and health data (Gulhane & Sajana, 2021), and a study on personality prediction

through social media expressions using MOET (Han et al., 2023).

All selected articles were published between 2016 and 2023, ensuring that the review

covers recent advancements in the field. No formal inclusion or exclusion criteria were applied

beyond relevance to the topic, publication quality, and availability in English.

 7
ML AND HUMAN BEHAVIOR PREDICTION ON SOCIAL MEDIA

Results and Discussion

The reviewed studies highlight the growing effectiveness of machine learning (ML)

techniques in analyzing and predicting human behavior on social media. Across all three selected

works, the ability of ML to process large volumes of user-generated data and extract meaningful

behavioral insights was a recurring theme.

Phan et al. (2016) introduced the SRBM+ model, which demonstrated that combining

user-level and community-level data improves both the accuracy and interpretability of

behavioral predictions. Their study emphasized the value of explainable AI, which is particularly

important in sensitive areas like health-related behavior prediction. This aligns with the growing

demand for transparent algorithms in both academic and clinical contexts.

Gulhane and Sajana (2021) offered a broader perspective by reviewing a variety of ML

techniques, including CNNs and regression models. Their findings revealed that combining

physiological data (e.g., EEG, ECG) with behavioral and social data enhances early disease

detection, such as for diabetes and heart conditions. This supports the idea that social media

behavior, when integrated with biometric signals, can serve as a reliable indicator of human well-

being.

Han et al. (2023) focused on personality prediction using social media text. By applying

multi-objective extra trees (MOET) and psycholinguistic dictionaries, the study successfully

linked language patterns to the Big Five personality traits. Importantly, it highlighted the

influence of culture through the use of the Culture Value Dictionary. This reveals how social

media users' expressions are shaped by cultural contexts, further enriching the interpretability of

ML predictions.
 8
ML AND HUMAN BEHAVIOR PREDICTION ON SOCIAL MEDIA

Taken together, the literature suggests that machine learning not only enables effective

behavioral predictions but also presents important challenges. These include data privacy risks,

algorithmic bias, and the complexity of interpreting model outputs. Despite these challenges, the

reviewed studies confirm that ML offers significant potential for applications in mental health

analysis, personality inference, and health behavior monitoring through social platforms.
 9
ML AND HUMAN BEHAVIOR PREDICTION ON SOCIAL MEDIA

Conclusion

This literature-guided review explored how machine learning (ML) is applied to predict

human behavior on social media. The study showed that machine learning techniques like neural

networks, and those ensemble models can actually pick up on complex things people express—

like emotions, thoughts, and even mental patterns—just from how they interact online.

However, there are still several challenges too. Things like data privacy, bias in the

algorithms, and how hard it is to actually understand what the models are doing—those are still

big concerns. If ML is going to be used more in studying human behavior, those problems really

need to be sorted out first.

In conclusion, the whole thing shows that machine learning (ML) is highly effective

when it comes to understanding people in online spaces. But it still needs to be used the right

way — with respect for privacy, more openness about how the systems work, and sensitivity to

different cultural backgrounds. If we really want to get the most out of this technology, we’ll

need more collaboration between different fields going forward.

 10
ML AND HUMAN BEHAVIOR PREDICTION ON SOCIAL MEDIA

References

Phan, N., Dou, D., Piniewski, B., & Kil, D. (2016). A deep learning approach for human

behavior prediction with explanations in health social networks: Social restricted

Boltzmann machine (SRBM+). Social Network Analysis and Mining, 6, Article 75.

https://doi.org/10.1007/s13278-016-0379-0

Gulhane, M., & Sajana, T. (2021). Human behavior prediction and analysis using machine

learning: A review. Turkish Journal of Computer and Mathematics Education, 12(5),

870–876. https://turcomat.org/index.php/turkbilmat/article/view/1499

Han, N., Li, S., Huang, F., Wen, Y., Su, Y., Li, L., Liu, X., & Zhu, T. (2023). How social media

expression can reveal personality. Frontiers in Psychiatry, 14, Article 1052844.

https://doi.org/10.3389/fpsyt.2023.1052844