Scholarlink

<p>Dengue fever poses a critical global health challenge, particularly in tropical and subtropical regions. Accurate identification of dengue-prone areas is essential for effective prevention and control. West Bengal, India, has witnessed significant dengue outbreaks, with the Kolkata-Howrah Municipal Corporation (KMC) area being the most affected. This study presents a novel approach, comparing Machine Learning (ML) and Multi-Criteria Decision Making (MCDM) techniques, to create a dengue susceptibility zonation model. We leverage diverse datasets, including environmental variables, demographic variables, and historical dengue incidence records collected through primary surveys. The models are built using the Fuzzy Analytic Hierarchy Process (F-AHP) and Random Forest (RF) algorithms, and their performance is assessed through Receiver Operating Characteristic (ROC) curve analysis, yielding Area Under the ROC Curve (AUC) values. While both models achieve similar AUC values, they produce different zonation patterns. To validate the models more accurately Mean Square Error (MSE), Root Mean Square Error (RMSE), Kappa Statistics have been employed. Random Forest generates a dispersed susceptibility map of KMC, while F-AHP yields a more concentrated pattern. Remarkably, Random Forest identifies “High-risk zones” effectively, with approximately 21.69% of dengue cases occurring in these areas, compared to 4.35% in the F-AHP model in KMC. Similarly, in HMC, 23.12% of area comes under “High-risk zones” in RF model while and 0.17% of area under the same zone in FAHP model. The result of MSE, RMSE and Kappa reveals the significance and accuracy of RF over F-AHP. Calculated values of MSE is higher in FAHP (0.085) than RF (0.070), followed by RMSE which shows similar validation where value of FAHP (0.291) is greater than RF (0.265) while the Kappa values of RF (0.913) is greater than FAHP (0.745). Thus, it can be stated that ML is giving much accurate result than MCDM approach. These zonation maps are invaluable for decision-makers, health authorities, and disaster management teams, aiding in proactive measures to curtail dengue spread and reduce mortality. This study underscores the power of ML techniques in predicting dengue susceptibility zones, reinforcing existing knowledge of dengue risk factors. By bridging advanced data analysis with practical applications, we advance dengue prevention and control efforts in West Bengal, offering insights that could benefit regions grappling with similar challenges worldwide.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Identification of dengue risk-prone areas using multicriteria decision-making model and machine learning algorithm in Kolkata and Howrah municipal corporation areas

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月14日

查看全文
  • Suvankar Naskar,
  • Brototi Biswas,
  • Jayanta Das,
  • Shuayb Abdinour Osman

摘要

Dengue fever poses a critical global health challenge, particularly in tropical and subtropical regions. Accurate identification of dengue-prone areas is essential for effective prevention and control. West Bengal, India, has witnessed significant dengue outbreaks, with the Kolkata-Howrah Municipal Corporation (KMC) area being the most affected. This study presents a novel approach, comparing Machine Learning (ML) and Multi-Criteria Decision Making (MCDM) techniques, to create a dengue susceptibility zonation model. We leverage diverse datasets, including environmental variables, demographic variables, and historical dengue incidence records collected through primary surveys. The models are built using the Fuzzy Analytic Hierarchy Process (F-AHP) and Random Forest (RF) algorithms, and their performance is assessed through Receiver Operating Characteristic (ROC) curve analysis, yielding Area Under the ROC Curve (AUC) values. While both models achieve similar AUC values, they produce different zonation patterns. To validate the models more accurately Mean Square Error (MSE), Root Mean Square Error (RMSE), Kappa Statistics have been employed. Random Forest generates a dispersed susceptibility map of KMC, while F-AHP yields a more concentrated pattern. Remarkably, Random Forest identifies “High-risk zones” effectively, with approximately 21.69% of dengue cases occurring in these areas, compared to 4.35% in the F-AHP model in KMC. Similarly, in HMC, 23.12% of area comes under “High-risk zones” in RF model while and 0.17% of area under the same zone in FAHP model. The result of MSE, RMSE and Kappa reveals the significance and accuracy of RF over F-AHP. Calculated values of MSE is higher in FAHP (0.085) than RF (0.070), followed by RMSE which shows similar validation where value of FAHP (0.291) is greater than RF (0.265) while the Kappa values of RF (0.913) is greater than FAHP (0.745). Thus, it can be stated that ML is giving much accurate result than MCDM approach. These zonation maps are invaluable for decision-makers, health authorities, and disaster management teams, aiding in proactive measures to curtail dengue spread and reduce mortality. This study underscores the power of ML techniques in predicting dengue susceptibility zones, reinforcing existing knowledge of dengue risk factors. By bridging advanced data analysis with practical applications, we advance dengue prevention and control efforts in West Bengal, offering insights that could benefit regions grappling with similar challenges worldwide.