Explainable assessment of financial experts' credibility by classifying social media forecasts and checking the predictions with actual market data

Silvia García-Méndez, Francisco de Arriba-Pérez, Jaime González-Gonzáleza, Francisco J. González-Castaño·June 17, 2024

Summary

This research paper presents a novel approach to evaluate the credibility of financial experts on social media by using Natural Language Processing (NLP) and Machine Learning (ML) to classify their asset value forecasts. The system assigns a continuous credibility score based on the accuracy of predictions compared to real market data, distinguishing it from binary classification methods. It analyzes social media metrics for correlation with credibility and provides model-agnostic explanations, enhancing transparency. The study contrasts existing research, which often treats credibility as binary, and aims to improve decision-making by offering reliable, real-time information with a focus on continuous assessment and explainability. The research also explores the use of various machine learning algorithms, feature engineering, and the impact of user context on credibility predictions. Future work includes refining the system, real-time training, and expanding to other domains like sports betting.

Key findings

3

Paper digest

What problem does the paper attempt to solve? Is this a new problem?

The paper aims to address the issue of identifying real-world credible experts in the financial domain by assessing their credibility through social media forecasts and validating these predictions with actual market data . This problem is not entirely new, as previous works have also focused on credibility assessment in various domains, including finance . However, the paper introduces a novel approach by classifying social media financial posts, validating them with market data, providing a credibility ranking on a continuous scale, and analyzing correlations with user context metrics for insights on audience interest in financial posts .


What scientific hypothesis does this paper seek to validate?

This paper aims to validate the scientific hypothesis related to assessing financial experts' credibility by classifying social media forecasts and verifying the predictions with actual market data . The study focuses on evaluating user credibility from social media financial posts, categorizing them into different types of forecasts, and then cross-referencing these forecasts with real market data to determine the credibility of financial experts . The research seeks to provide a continuous credibility ranking scale based on the analysis of user context metrics and aims to establish correlations between credibility ranking and user context to gain insights into audience interest in financial posts . Additionally, the paper explores the use of natural language explainability techniques to describe model predictions, enhancing the transparency and interpretability of the assessment process .


What new ideas, methods, or models does the paper propose? What are the characteristics and advantages compared to previous methods?

The paper proposes a comprehensive solution for the credibility assessment of finance-oriented users, incorporating innovative ideas, methods, and models . Here are the key aspects of the proposed methodology:

  1. Pre-processing Module: The pre-processing module focuses on preparing textual content by removing irrelevant data such as links to external sources, images, stop words, special characters, and emoticons. It also includes lemmatization to streamline the text for further analysis .

  2. Feature Engineering Module: This module generates n-grams and side features to enhance the analysis. It incorporates content-based features like readability score, reading time, number of complex words, emotion, and polarity data. Additionally, textual features such as char and word n-grams are utilized to extract valuable information from the text .

  3. Classification Module: The hybrid classification module combines term-based detection and machine learning (ML) prediction to categorize posts into types of forecasts. The forecast categories considered are short-term drops, short-term rises, and a third category for other posts mentioning stock values. The system uses a term lexicon for each target category and employs supervised ML models trained with specific features for accurate classification .

  4. Quality Assessment Module: To evaluate the predictions, the system compares the predicted forecasts with the actual evolution of stock values. The quality rank is determined based on the number of successful predictions, focusing on posts mentioning stocks or markets for evaluation .

  5. Explainability Module: The explainability module plays a crucial role in enhancing the trustworthiness of the automatic analysis. It provides explanations in natural language based on relevant terms or features extracted during the classification process. The system employs a model-agnostic approach to ensure transparency and trust in the analysis results .

Overall, the paper introduces a systematic approach that integrates pre-processing, feature engineering, classification, quality assessment, and explainability modules to assess the credibility of financial experts based on social media forecasts and actual market data . The proposed methodology for assessing the credibility of finance-oriented users through social media forecasts offers several distinct characteristics and advantages compared to previous methods :

  1. Comprehensive Approach: The methodology integrates pre-processing, feature engineering, classification, quality assessment, and explainability modules to provide a holistic solution for credibility assessment .

  2. Pre-processing Module: The methodology includes a pre-processing module that removes irrelevant data such as links, images, stop words, special characters, and emoticons, ensuring optimal knowledge extraction from textual content .

  3. Feature Engineering Module: It focuses on generating n-grams, side features, and content-based features like readability score, reading time, emotion, and polarity data, enhancing the analysis of social media forecasts .

  4. Hybrid Classification Module: The methodology combines term-based detection and machine learning (ML) prediction to categorize posts into types of forecasts, improving efficiency and accuracy in classification .

  5. Quality Assessment Module: The system ranks prediction quality by comparing forecasts with actual market data, providing a reliable evaluation metric for the credibility assessment of financial experts .

  6. Explainability Module: The methodology incorporates an explainability module that generates natural language descriptions of decisions, enhancing transparency and trust in the analysis results. This feature empowers non-expert investors to understand the classifier's decisions and make informed investment choices .

  7. Continuous Ranking: Unlike previous binary classification approaches, the proposed methodology provides a continuous credibility ranking scale, offering a more nuanced assessment of financial experts' credibility .

  8. Real-world Validation: The methodology validates social media forecasts with actual market data, ensuring the practical applicability and accuracy of the credibility assessment .

  9. Feature Relevance: The methodology leverages feature relevance to create natural language descriptions, enhancing the interpretability of model predictions and providing valuable insights for users .

  10. Adaptability and Future Enhancements: The methodology plans to explore real-time data training, reinforcement learning, and adaptive precision methods to further enhance the system's adaptability and performance. It also aims to apply the approach to other fields beyond finance, such as online gaming, showcasing its versatility and potential for broader applications .


Do any related researches exist? Who are the noteworthy researchers on this topic in this field?What is the key to the solution mentioned in the paper?

Several related research studies exist in the field of assessing financial experts' credibility by classifying social media forecasts and validating predictions with actual market data. Noteworthy researchers in this field include Huang, T.-C., Zaeem, R. N., Barber, K. S., Hudon, A., Beaudoin, M., Phraxayavong, K., Jalal, N., Mehmood, A., Choi, G. S., Ashraf, I., Verma, P. K., Agrawal, P., Madaan, V., Gupta, C., and many others .

The key to the solution mentioned in the paper involves a multi-faceted approach that includes:

  • Assessing user credibility from social media financial posts that are automatically classified into types of forecasts and validated with actual market data.
  • Providing a credibility ranking as a continuous scale.
  • Analyzing correlations between credibility ranking and user context metrics to understand audience interest in financial posts.
  • Describing model predictions with natural language explainability techniques.
  • Utilizing a model-agnostic approach to extract relevant features for explanations in natural language.
  • Offering credibility ranking on a continuous scale, differentiating it from other existing systems in the literature .

How were the experiments in the paper designed?

The experiments in the paper were designed as follows:

  • The experimental data set consisted of 15,000 tweets from 12 Spanish trading advisors, with 1000 posts per advisor, published between January 16, 2017, and June 5, 2023. These tweets were manually annotated into three categories: short-term drop, short-term rise, and other, by financial experts to ensure diversity in Twitter statistics .
  • Term-based detection was used to predict 62.99% of the samples in the experimental data set, which helped save computing resources and time. Subsequently, ml classification algorithms were evaluated using 10-fold cross-validation .
  • The experiments involved the use of various machine learning algorithms such as Linear Support Vector Machines (lsvm), Decision Trees (dt), k-Nearest Neighbors (knn), Naive Bayes (nb), Random Forest (rf), and Gradient Boosting (gb) to classify the posts into different forecast categories .
  • The quality assessment module compared the predicted forecasts with the actual stock values' evolution, identified with $cashtags and financial tickers, to determine the quality of predictions. Only posts mentioning stocks or markets were considered for quality evaluation .
  • The explainability module provided explanations in natural language based on relevant terms or features to enhance the trustworthiness of the automatic analysis. It presented explanations for posts classified by term-based detection and ml models using a model-agnostic approach .

What is the dataset used for quantitative evaluation? Is the code open source?

The dataset used for quantitative evaluation in the study is a set of 15,000 tweets published by 12 Spanish trading advisors between January 16, 2017, and June 5, 2023. These tweets were manually annotated by financial experts into three categories: short-term drop, short-term rise, and other . The code for the machine learning classification algorithms used in the study is open source and implemented in Python using the scikit-learn library .


Do the experiments and results in the paper provide good support for the scientific hypotheses that need to be verified? Please analyze.

The experiments and results presented in the paper provide strong support for the scientific hypotheses that needed verification. The study conducted experiments using a dataset of 15,000 tweets from 12 Spanish trading advisors, which were manually annotated into different categories by financial experts . The experiments involved classifying social media forecasts, checking predictions with actual market data, and assessing the credibility of financial experts . This methodology allowed for a comprehensive analysis of financial experts' credibility based on social media content and real-world market outcomes.

Furthermore, the paper's approach included assessing user credibility from social media financial posts, providing a credibility ranking on a continuous scale, and analyzing correlations between credibility ranking and user context metrics . This multi-faceted analysis enhances the robustness of the study's findings and supports the scientific hypotheses under investigation.

Moreover, the results obtained from the experiments were based on a detailed evaluation of the experimental data set, the implementation of modules, and the outcomes achieved . The study's methodology, which involved pre-processing modules to identify specific elements in the tweets, contributed to the reliability and validity of the results obtained.

Overall, the experiments and results presented in the paper offer substantial evidence to support the scientific hypotheses related to the assessment of financial experts' credibility through the classification of social media forecasts and validation with actual market data. The comprehensive approach taken in the study, along with the detailed analysis of the experimental data set and results, strengthens the credibility and reliability of the findings .


What are the contributions of this paper?

The paper makes several contributions:

  • It identifies real-world credible experts in the financial domain .
  • It implements a machine learning algorithm for automated thematic annotations in Avatar using a linear support vector classifier approach .
  • It presents a novel improved random forest for text classification using feature ranking and optimal number of trees .
  • It evaluates information credibility in online professional social networks using a tree augmented naïve Bayes classifier .
  • It introduces VAPER, a deep learning model for explainable probabilistic regression .
  • It enhances fake news detection through DeepFakE using tensor decomposition-based deep neural networks .
  • It fuses machine learning and deep learning methods for user credibility on social media in the UCred system .

What work can be continued in depth?

Further research in the field of financial experts' credibility assessment can be expanded in several directions based on the existing work:

  • Enhancing User Context Analysis: Future studies can delve deeper into analyzing user context metrics to gain more insights into the audience's interests in financial posts. This can involve exploring additional user-related factors that may influence credibility rankings and decision-making processes .
  • Improving Explainability Techniques: There is room for advancement in developing more sophisticated natural language explanations for model predictions. Research can focus on enhancing the clarity and depth of these explanations to further increase the trustworthiness of the automatic analysis .
  • Exploring New ML Models: Investigating the effectiveness of other machine learning models beyond the ones commonly used, such as Linear Support Vector Machines, Decision Trees, and Random Forest, could be a valuable area for future research. Exploring novel models may lead to improved accuracy and efficiency in classifying financial posts and assessing credibility .
  • Real-time Prediction Optimization: While the proposed solution performs real-time predictions, there is potential for optimizing the training process to reduce time consumption compared to rule-based approaches. Future research could focus on streamlining the training phase while maintaining the real-time prediction capability .
  • Incorporating Additional Features: Researchers can consider incorporating new features related to social media posts, user behavior, or market data to enhance the credibility assessment process. By expanding the feature set, the model's predictive power and accuracy could be further improved .

Introduction
Background
Evolution of social media as a source of financial information
Importance of expert credibility in financial decision-making
Objective
To develop a novel credibility assessment system
Improve upon binary classification methods
Enhance transparency and real-time decision-making
Methodology
Data Collection
Social media data from financial experts
Asset value forecast data
Social media metrics (likes, shares, engagement)
Data Preprocessing
Text analysis using NLP techniques
Data cleaning and normalization
Feature extraction from social media data
Model Development
Natural Language Processing
Sentiment analysis
Topic modeling
Entity recognition
Machine Learning Algorithms
Comparison of algorithms (e.g., SVM, Random Forest, LSTM)
Model selection based on performance
Feature Engineering
Creation of credibility-related features
Integration of social media metrics
Continuous Assessment
Real-time prediction and updating of credibility scores
Explainability
Model-agnostic explanations for credibility predictions
Evaluation and Comparison
Binary vs. continuous credibility classification
Performance metrics (accuracy, precision, recall)
Contrast with existing research methods
Results and Discussion
Credibility score distribution
Impact of user context on credibility predictions
Limitations and challenges faced
Future Work
Refinement of the credibility assessment system
Real-time training and adaptation
Application to sports betting and other domains
Conclusion
Summary of key findings
Implications for financial decision-making and social media analysis
Suggestions for future research directions
Basic info
papers
computation and language
machine learning
social and information networks
artificial intelligence
Advanced features
Insights
What method does the research paper propose for evaluating financial experts' credibility on social media?
What is the primary focus of the study in terms of improving decision-making for users?
How does the system differ from binary classification methods in assessing credibility?
What are the key components of the system that enhance transparency, according to the research?

Explainable assessment of financial experts' credibility by classifying social media forecasts and checking the predictions with actual market data

Silvia García-Méndez, Francisco de Arriba-Pérez, Jaime González-Gonzáleza, Francisco J. González-Castaño·June 17, 2024

Summary

This research paper presents a novel approach to evaluate the credibility of financial experts on social media by using Natural Language Processing (NLP) and Machine Learning (ML) to classify their asset value forecasts. The system assigns a continuous credibility score based on the accuracy of predictions compared to real market data, distinguishing it from binary classification methods. It analyzes social media metrics for correlation with credibility and provides model-agnostic explanations, enhancing transparency. The study contrasts existing research, which often treats credibility as binary, and aims to improve decision-making by offering reliable, real-time information with a focus on continuous assessment and explainability. The research also explores the use of various machine learning algorithms, feature engineering, and the impact of user context on credibility predictions. Future work includes refining the system, real-time training, and expanding to other domains like sports betting.
Mind map
Model-agnostic explanations for credibility predictions
Real-time prediction and updating of credibility scores
Integration of social media metrics
Creation of credibility-related features
Model selection based on performance
Comparison of algorithms (e.g., SVM, Random Forest, LSTM)
Entity recognition
Topic modeling
Sentiment analysis
Explainability
Continuous Assessment
Feature Engineering
Machine Learning Algorithms
Natural Language Processing
Feature extraction from social media data
Data cleaning and normalization
Text analysis using NLP techniques
Social media metrics (likes, shares, engagement)
Asset value forecast data
Social media data from financial experts
Enhance transparency and real-time decision-making
Improve upon binary classification methods
To develop a novel credibility assessment system
Importance of expert credibility in financial decision-making
Evolution of social media as a source of financial information
Suggestions for future research directions
Implications for financial decision-making and social media analysis
Summary of key findings
Application to sports betting and other domains
Real-time training and adaptation
Refinement of the credibility assessment system
Limitations and challenges faced
Impact of user context on credibility predictions
Credibility score distribution
Contrast with existing research methods
Performance metrics (accuracy, precision, recall)
Binary vs. continuous credibility classification
Model Development
Data Preprocessing
Data Collection
Objective
Background
Conclusion
Future Work
Results and Discussion
Evaluation and Comparison
Methodology
Introduction
Outline
Introduction
Background
Evolution of social media as a source of financial information
Importance of expert credibility in financial decision-making
Objective
To develop a novel credibility assessment system
Improve upon binary classification methods
Enhance transparency and real-time decision-making
Methodology
Data Collection
Social media data from financial experts
Asset value forecast data
Social media metrics (likes, shares, engagement)
Data Preprocessing
Text analysis using NLP techniques
Data cleaning and normalization
Feature extraction from social media data
Model Development
Natural Language Processing
Sentiment analysis
Topic modeling
Entity recognition
Machine Learning Algorithms
Comparison of algorithms (e.g., SVM, Random Forest, LSTM)
Model selection based on performance
Feature Engineering
Creation of credibility-related features
Integration of social media metrics
Continuous Assessment
Real-time prediction and updating of credibility scores
Explainability
Model-agnostic explanations for credibility predictions
Evaluation and Comparison
Binary vs. continuous credibility classification
Performance metrics (accuracy, precision, recall)
Contrast with existing research methods
Results and Discussion
Credibility score distribution
Impact of user context on credibility predictions
Limitations and challenges faced
Future Work
Refinement of the credibility assessment system
Real-time training and adaptation
Application to sports betting and other domains
Conclusion
Summary of key findings
Implications for financial decision-making and social media analysis
Suggestions for future research directions
Key findings
3

Paper digest

What problem does the paper attempt to solve? Is this a new problem?

The paper aims to address the issue of identifying real-world credible experts in the financial domain by assessing their credibility through social media forecasts and validating these predictions with actual market data . This problem is not entirely new, as previous works have also focused on credibility assessment in various domains, including finance . However, the paper introduces a novel approach by classifying social media financial posts, validating them with market data, providing a credibility ranking on a continuous scale, and analyzing correlations with user context metrics for insights on audience interest in financial posts .


What scientific hypothesis does this paper seek to validate?

This paper aims to validate the scientific hypothesis related to assessing financial experts' credibility by classifying social media forecasts and verifying the predictions with actual market data . The study focuses on evaluating user credibility from social media financial posts, categorizing them into different types of forecasts, and then cross-referencing these forecasts with real market data to determine the credibility of financial experts . The research seeks to provide a continuous credibility ranking scale based on the analysis of user context metrics and aims to establish correlations between credibility ranking and user context to gain insights into audience interest in financial posts . Additionally, the paper explores the use of natural language explainability techniques to describe model predictions, enhancing the transparency and interpretability of the assessment process .


What new ideas, methods, or models does the paper propose? What are the characteristics and advantages compared to previous methods?

The paper proposes a comprehensive solution for the credibility assessment of finance-oriented users, incorporating innovative ideas, methods, and models . Here are the key aspects of the proposed methodology:

  1. Pre-processing Module: The pre-processing module focuses on preparing textual content by removing irrelevant data such as links to external sources, images, stop words, special characters, and emoticons. It also includes lemmatization to streamline the text for further analysis .

  2. Feature Engineering Module: This module generates n-grams and side features to enhance the analysis. It incorporates content-based features like readability score, reading time, number of complex words, emotion, and polarity data. Additionally, textual features such as char and word n-grams are utilized to extract valuable information from the text .

  3. Classification Module: The hybrid classification module combines term-based detection and machine learning (ML) prediction to categorize posts into types of forecasts. The forecast categories considered are short-term drops, short-term rises, and a third category for other posts mentioning stock values. The system uses a term lexicon for each target category and employs supervised ML models trained with specific features for accurate classification .

  4. Quality Assessment Module: To evaluate the predictions, the system compares the predicted forecasts with the actual evolution of stock values. The quality rank is determined based on the number of successful predictions, focusing on posts mentioning stocks or markets for evaluation .

  5. Explainability Module: The explainability module plays a crucial role in enhancing the trustworthiness of the automatic analysis. It provides explanations in natural language based on relevant terms or features extracted during the classification process. The system employs a model-agnostic approach to ensure transparency and trust in the analysis results .

Overall, the paper introduces a systematic approach that integrates pre-processing, feature engineering, classification, quality assessment, and explainability modules to assess the credibility of financial experts based on social media forecasts and actual market data . The proposed methodology for assessing the credibility of finance-oriented users through social media forecasts offers several distinct characteristics and advantages compared to previous methods :

  1. Comprehensive Approach: The methodology integrates pre-processing, feature engineering, classification, quality assessment, and explainability modules to provide a holistic solution for credibility assessment .

  2. Pre-processing Module: The methodology includes a pre-processing module that removes irrelevant data such as links, images, stop words, special characters, and emoticons, ensuring optimal knowledge extraction from textual content .

  3. Feature Engineering Module: It focuses on generating n-grams, side features, and content-based features like readability score, reading time, emotion, and polarity data, enhancing the analysis of social media forecasts .

  4. Hybrid Classification Module: The methodology combines term-based detection and machine learning (ML) prediction to categorize posts into types of forecasts, improving efficiency and accuracy in classification .

  5. Quality Assessment Module: The system ranks prediction quality by comparing forecasts with actual market data, providing a reliable evaluation metric for the credibility assessment of financial experts .

  6. Explainability Module: The methodology incorporates an explainability module that generates natural language descriptions of decisions, enhancing transparency and trust in the analysis results. This feature empowers non-expert investors to understand the classifier's decisions and make informed investment choices .

  7. Continuous Ranking: Unlike previous binary classification approaches, the proposed methodology provides a continuous credibility ranking scale, offering a more nuanced assessment of financial experts' credibility .

  8. Real-world Validation: The methodology validates social media forecasts with actual market data, ensuring the practical applicability and accuracy of the credibility assessment .

  9. Feature Relevance: The methodology leverages feature relevance to create natural language descriptions, enhancing the interpretability of model predictions and providing valuable insights for users .

  10. Adaptability and Future Enhancements: The methodology plans to explore real-time data training, reinforcement learning, and adaptive precision methods to further enhance the system's adaptability and performance. It also aims to apply the approach to other fields beyond finance, such as online gaming, showcasing its versatility and potential for broader applications .


Do any related researches exist? Who are the noteworthy researchers on this topic in this field?What is the key to the solution mentioned in the paper?

Several related research studies exist in the field of assessing financial experts' credibility by classifying social media forecasts and validating predictions with actual market data. Noteworthy researchers in this field include Huang, T.-C., Zaeem, R. N., Barber, K. S., Hudon, A., Beaudoin, M., Phraxayavong, K., Jalal, N., Mehmood, A., Choi, G. S., Ashraf, I., Verma, P. K., Agrawal, P., Madaan, V., Gupta, C., and many others .

The key to the solution mentioned in the paper involves a multi-faceted approach that includes:

  • Assessing user credibility from social media financial posts that are automatically classified into types of forecasts and validated with actual market data.
  • Providing a credibility ranking as a continuous scale.
  • Analyzing correlations between credibility ranking and user context metrics to understand audience interest in financial posts.
  • Describing model predictions with natural language explainability techniques.
  • Utilizing a model-agnostic approach to extract relevant features for explanations in natural language.
  • Offering credibility ranking on a continuous scale, differentiating it from other existing systems in the literature .

How were the experiments in the paper designed?

The experiments in the paper were designed as follows:

  • The experimental data set consisted of 15,000 tweets from 12 Spanish trading advisors, with 1000 posts per advisor, published between January 16, 2017, and June 5, 2023. These tweets were manually annotated into three categories: short-term drop, short-term rise, and other, by financial experts to ensure diversity in Twitter statistics .
  • Term-based detection was used to predict 62.99% of the samples in the experimental data set, which helped save computing resources and time. Subsequently, ml classification algorithms were evaluated using 10-fold cross-validation .
  • The experiments involved the use of various machine learning algorithms such as Linear Support Vector Machines (lsvm), Decision Trees (dt), k-Nearest Neighbors (knn), Naive Bayes (nb), Random Forest (rf), and Gradient Boosting (gb) to classify the posts into different forecast categories .
  • The quality assessment module compared the predicted forecasts with the actual stock values' evolution, identified with $cashtags and financial tickers, to determine the quality of predictions. Only posts mentioning stocks or markets were considered for quality evaluation .
  • The explainability module provided explanations in natural language based on relevant terms or features to enhance the trustworthiness of the automatic analysis. It presented explanations for posts classified by term-based detection and ml models using a model-agnostic approach .

What is the dataset used for quantitative evaluation? Is the code open source?

The dataset used for quantitative evaluation in the study is a set of 15,000 tweets published by 12 Spanish trading advisors between January 16, 2017, and June 5, 2023. These tweets were manually annotated by financial experts into three categories: short-term drop, short-term rise, and other . The code for the machine learning classification algorithms used in the study is open source and implemented in Python using the scikit-learn library .


Do the experiments and results in the paper provide good support for the scientific hypotheses that need to be verified? Please analyze.

The experiments and results presented in the paper provide strong support for the scientific hypotheses that needed verification. The study conducted experiments using a dataset of 15,000 tweets from 12 Spanish trading advisors, which were manually annotated into different categories by financial experts . The experiments involved classifying social media forecasts, checking predictions with actual market data, and assessing the credibility of financial experts . This methodology allowed for a comprehensive analysis of financial experts' credibility based on social media content and real-world market outcomes.

Furthermore, the paper's approach included assessing user credibility from social media financial posts, providing a credibility ranking on a continuous scale, and analyzing correlations between credibility ranking and user context metrics . This multi-faceted analysis enhances the robustness of the study's findings and supports the scientific hypotheses under investigation.

Moreover, the results obtained from the experiments were based on a detailed evaluation of the experimental data set, the implementation of modules, and the outcomes achieved . The study's methodology, which involved pre-processing modules to identify specific elements in the tweets, contributed to the reliability and validity of the results obtained.

Overall, the experiments and results presented in the paper offer substantial evidence to support the scientific hypotheses related to the assessment of financial experts' credibility through the classification of social media forecasts and validation with actual market data. The comprehensive approach taken in the study, along with the detailed analysis of the experimental data set and results, strengthens the credibility and reliability of the findings .


What are the contributions of this paper?

The paper makes several contributions:

  • It identifies real-world credible experts in the financial domain .
  • It implements a machine learning algorithm for automated thematic annotations in Avatar using a linear support vector classifier approach .
  • It presents a novel improved random forest for text classification using feature ranking and optimal number of trees .
  • It evaluates information credibility in online professional social networks using a tree augmented naïve Bayes classifier .
  • It introduces VAPER, a deep learning model for explainable probabilistic regression .
  • It enhances fake news detection through DeepFakE using tensor decomposition-based deep neural networks .
  • It fuses machine learning and deep learning methods for user credibility on social media in the UCred system .

What work can be continued in depth?

Further research in the field of financial experts' credibility assessment can be expanded in several directions based on the existing work:

  • Enhancing User Context Analysis: Future studies can delve deeper into analyzing user context metrics to gain more insights into the audience's interests in financial posts. This can involve exploring additional user-related factors that may influence credibility rankings and decision-making processes .
  • Improving Explainability Techniques: There is room for advancement in developing more sophisticated natural language explanations for model predictions. Research can focus on enhancing the clarity and depth of these explanations to further increase the trustworthiness of the automatic analysis .
  • Exploring New ML Models: Investigating the effectiveness of other machine learning models beyond the ones commonly used, such as Linear Support Vector Machines, Decision Trees, and Random Forest, could be a valuable area for future research. Exploring novel models may lead to improved accuracy and efficiency in classifying financial posts and assessing credibility .
  • Real-time Prediction Optimization: While the proposed solution performs real-time predictions, there is potential for optimizing the training process to reduce time consumption compared to rule-based approaches. Future research could focus on streamlining the training phase while maintaining the real-time prediction capability .
  • Incorporating Additional Features: Researchers can consider incorporating new features related to social media posts, user behavior, or market data to enhance the credibility assessment process. By expanding the feature set, the model's predictive power and accuracy could be further improved .
Scan the QR code to ask more questions about the paper
© 2025 Powerdrill. All rights reserved.