Machine Learning for HIV-ART Optimization
Abstract:
Bibliography/Citations:
- https://www.google.com/url?sa=i&url=https%3A%2F%2Fmljar.com%2Fblog%2Fxgboost-early- stopping%2F&psig=AOvVaw1HgwO2jJRsPR6yFMVM0D9D&ust=1708214474676000&source=images&cd=vfe&opi=89978449&ved=0CBMQjRxqFwoTCLCYhvqlsYQDFQAAAAADAAAAABAD
- https://www.google.com/url?sa=i&url=https%3A%2F%2Fourworldindata.org%2Fgrapher%2Fhiv-deaths-averted-from-antiretroviral-therapy-art&psig=AOvVaw1Z67YWDxcq246pgrzdu UkJ&ust=1708214404053000&source=images&cd=vfe&opi=89978449&ved=0CBMQjRxqFwoTCNDOsruOsYQDFQAAAAADAAAAABAD
- https://www.google.com/url?sa=i&url=https%3A%2F%2Fwww.viewsoftheworld.net%2F%3Fp%3D5015&psig=AOvVaw35AIGu3quXPgXtrpq6Vy_v&ust=1708214300142000&source=images&cd=vfe&opi=89978449&ved=0CBMQjRxqFwoTCMD5v8WOSYQDFQAAAAAAAAAABAD
- "Hiv around the World - Google Search." Google.com, 2019, www.google.com/search?sca_esv=c33fa338858c7c86&q=hiv+around+the+world&tbm=isch&source=lnms&sa=X&ved=2ahUKEwjC3- SjiLGEAXWD120FHXaOCOYQ0pQJegQICxAB#imgrc=exSODpiB47aPLM. Accessed 16 Feb. 2024.
- "Light Gradient Boosted Trees Regressor with Early Stopping - Google Search." Google.com, 2019, www.google.com/search?q light+gradient-boosted+trees+regressor+with+early+stopping+&tbm=isch&ved 2ahUKEwjdsqbXiLGEAxWrzekDHbb2BIkQ2- cCcgQIABAA&oq=light+gradient+boosted+trees-regressor+with+carly+stopping+&gs_lp=EgNpbWciO2xpZ2h0IGdyYWRpZW50IGJvb3N0ZWQgdHJ1ZXMgcmVncmVzc29yIHdpdGggZWFybHkgc3RvcHBpbmcgSPZmUJgPWIFjcAF4AJABAZgBzwKgAZE3qgEJMTQuNDUuMS4xuAEDYAEA-AEBigILZ3dzLXdpeil pbWfCAgUQABIABMICCBAAGIAEGLEDwgIKEAAYgAQYigUYQ8ICDRAAGIAEGIOFGEMYSQPCAgeQABIABBgYiAYB&sclient=img&ei=BPfPZd2gA6ubp84Ptu2TyAg#imgrc 16Xs_aDqV0Ot8M. Accessed 17 Feb. 2024.
- "Number of Hiv Deaths Averted from Antiretroviral Therapy (ART), 2022 - Google Search." Google.com, 2022, www.google.com/search?q=number+of+hiv+deaths+averted+from+antiretroviral therapy (ART) %2C+2022&tbm=isch&ved=2ahUKEwi21OCliLGEAxUwHNAFHS9VB3gQ2-cCegQIABAA&oq=number+of+hiv+deaths averted+from+antiretroviral therapy(ART)%2C+2022&gs_lp=EgNpbWciQ25 lbWJlciBvZiBoaXYgZGVhdGhzIGF2ZXJ0ZWQgZnJvbSBhbnRpcmV0cm92aXJhbCB0aGVyYXB5KEFSVCksIDIwMjJI384BUMcHWMXGAXALACQAQGYAfwDoAG4TaoBCzEwLjYzLjMuNS0xuAEDYAEA-AEBigILZ3dzLXdpeil pbWICAgoQABIABBIKBRhDwgIGEAAYCBgewgIHEAAYgAQYGMICCBAAGIAEGLEDwgIFEAAYgATCAg0QABIABBIKBRIDGLEDIAYB&sclient=img&ei=nPbPZbbWCLC4wN4Pr6qdwAc#imgrc=N-6uOHRPI7ZnLM. Accessed 17 Feb. 2024.
Additional Project Information
Research Plan:
A. Question or Problem being addressed: What are the areas in need of antiretroviral therapy (ART) treatment for children with HIV, and to what extent do these countries influence the global HIV landscape? How can machine learning models be used to predict the need for ART treatment and assess the global impact of these countries?
B. Goals
- Develop machine learning models to predict which countries require ART treatment for children with HIV.
- Assess the global influence of these countries on the HIV landscape.
- Investigate the relationship between healthcare discrepancies and the prevalence of HIV in children.
- Test the hypothesis that increased healthcare is more essential to underdeveloped countries than developed ones in addressing the HIV epidemic among children.
C. Procedure
Data Collection: a. Gather data on HIV prevalence, ART coverage, healthcare infrastructure, and socioeconomic indicators from reputable sources such as hiv.gov. b. Collect data spanning from 2010 to 2022 to capture trends over time. c. Ensure data quality and consistency across different regions.
Model Development: a. Utilize AI machine learning platforms such as DataRobot for model development. b. Implement features such as user-defined grouping and RMSE optimization metric to enhance model performance. c. Train machine learning models using regression techniques and ensemble methods. d. Conduct 10 to 15 trials for model training and accuracy assessment. e. Refine the models through iterative trials and adjustments based on performance metrics.
Data Analysis:
Exploratory Data Analysis: a. Analyze the distribution and relationships between variables using descriptive statistics and data visualization techniques. b. Identify patterns and trends in the data related to HIV prevalence, ART coverage, and healthcare infrastructure.
Model Evaluation: a. Assess the performance of the machine learning models using cross-validation techniques. b. Calculate prediction percentages for each country indicating the need for ART treatment and its influence on the global HIV landscape. c. Validate the models using lift charts to visualize their effectiveness in segmenting the target population and predicting the need for ART treatment.
Statistical Analysis: a. Conduct statistical tests to investigate the relationship between healthcare discrepancies and HIV prevalence in children. b. Test hypotheses regarding the differential impact of healthcare on HIV outcomes between developed and underdeveloped countries.
Interpretation of Results: a. Interpret the results to identify countries in need of ART treatment and assess their global influence on the HIV landscape. b. Draw conclusions regarding the efficacy of predictive modeling in healthcare resource allocation and policy decisions. c. Discuss the implications of the findings for addressing the HIV epidemic among children in developed and underdeveloped countries.
This research plan outlines a systematic approach to developing machine learning models for predicting the need for ART treatment in children with HIV and assessing the global influence of these countries. It includes procedures for data collection, model development, data analysis, and interpretation of results to address the research questions and hypotheses outlined in the abstract.
Questions and Answers
1.The major objective of the project was to identify areas in need of antiretroviral therapy (ART) treatment for children with HIV and to assess the global influence of these countries, achieved by utilizing AI machine learning models to predict which countries required ART treatment and to what extent they influenced the global HIV landscape. This goal stemmed from the pressing issue of healthcare discrepancies between developed and underdeveloped nations, particularly in the context of HIV treatment for children. The objective wasn't a result of any specific situation or experience but aimed to solve the problem of efficiently allocating healthcare resources by testing the hypothesis that increased healthcare is more essential to underdeveloped countries.
2.The major tasks included gathering and preprocessing the necessary data related to HIV prevalence, ART coverage, and global influence, utilizing AI machine learning models, specifically DataRobot, to train predictive models and assess the need for ART treatment in different countries, and implementing various features and optimization techniques, such as user-defined grouping and RMSE optimization metric, to enhance the accuracy of the models. For teams, each member may have focused on specific aspects such as data gathering, model development, or result interpretation.
3.The novelty of the project lies in its use of AI machine learning models to predict the need for ART treatment in children with HIV on a global scale, utilizing DataRobot and implementing features like user-defined grouping and RMSE optimization metric. This approach may not have been done before in this specific context, representing a unique approach to addressing healthcare discrepancies in HIV treatment for children. Research was conducted to ensure that the methodology and objectives of the project were unique and aligned with the current state of research in the field of HIV treatment and healthcare disparities.
4.The most challenging part of completing the project was likely overcoming the complexities of working with healthcare data, particularly in the context of global health disparities and HIV treatment. Problems encountered may have included data inconsistencies, model optimization, and interpreting the results accurately. To overcome these challenges, thorough data preprocessing and validation procedures were likely implemented, along with iterative model refinement and validation, providing valuable insights into working with healthcare data and machine learning models, enhancing the understanding of healthcare disparities and the effectiveness of predictive modeling in this context.
5.If the project were to be done again, there may be several things done differently, such as exploring additional features or data sources to improve the accuracy and robustness of the predictive models, conducting further analysis on the socioeconomic factors influencing healthcare disparities and HIV treatment outcomes, and collaborating with healthcare professionals and policymakers to ensure the practical applicability of the project's findings.
6.Working on this project may have sparked ideas for other related projects, such as investigating the effectiveness of different healthcare interventions in reducing HIV prevalence among children in underdeveloped countries, developing decision support systems for healthcare resource allocation based on predictive modeling and machine learning techniques, and exploring the impact of healthcare policies and interventions on global health outcomes, particularly in the context of infectious diseases like HIV/AIDS.
7.COVID-19 may have affected the completion of the project in various ways, such as disruption of data collection and processing procedures due to lockdowns and restrictions on research activities, shifts in priorities and resources within healthcare systems, potentially affecting the availability of data related to HIV treatment and healthcare disparities, and changes in the global health landscape and policy responses may have influenced the interpretation and implications of the project's findings.