Scholarlink

<p>Speed dating is a social gathering where attendees have a series of quick one-on-one meetings with possible love partners. These dates, which are usually just a few minutes long, enable people to rapidly determine compatibility and shared interests. Participants mark the people they would like to see again following the event. If interest is mutual, contact information is exchanged for future follow-up conversations and possible dates. However, matching success remains inconsistent due to the subjective nature of these brief encounters. Through the use of participant interaction data to forecast results, machine learning (ML) improves speed dating. This predictive feature improves matching accuracy by forecasting relationship success using past data. In this study, the performance of two Extreme Learning Machine (ELM) and Support Vector Classification (SVC) base models was cross-validated using a compiled dataset. To further enhance predictive models, optimization techniques like the Mother Optimization Algorithm (MOA) and Stochastic Paint Optimizer are used. Also, feature importance analysis and the Wilcoxon test were utilized to evaluate the significance of individual features and to statistically compare the performance of the models, confirming that the improvements achieved by the optimized models were statistically significant. Results indicate that these optimizations also greatly enhanced model performance. For instance, ELMA (MOA-optimized ELM) outperformed the baseline ELM model by boosting test accuracy from 91.29 to 94.52% and precision from 91.84 to 94.56%. Similarly, SVCA (MOA-optimized SVC) achieved a 93.55% test accuracy, higher than the baseline SVC's 90.00%. Overall, ELMA delivered the best performance, with precision scores of 0.94 and 0.92 in mismatched and matched cases, while the base SVC showed the weakest results, with 0.98 and 0.70 precision in those conditions. Integrating optimization algorithms with ML models substantially boosts the accuracy of predicting successful matches in speed dating, offering a promising approach to improving dating outcomes.</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

Enhancing speed dating predictions using machine learning approaches and models performance analysis

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月17日

查看全文
  • Dongmei Guo,
  • Jinying Wang,
  • Haifeng Ma,
  • Chengnan Li,
  • Lin Liu,
  • Wangyan Li

摘要

Speed dating is a social gathering where attendees have a series of quick one-on-one meetings with possible love partners. These dates, which are usually just a few minutes long, enable people to rapidly determine compatibility and shared interests. Participants mark the people they would like to see again following the event. If interest is mutual, contact information is exchanged for future follow-up conversations and possible dates. However, matching success remains inconsistent due to the subjective nature of these brief encounters. Through the use of participant interaction data to forecast results, machine learning (ML) improves speed dating. This predictive feature improves matching accuracy by forecasting relationship success using past data. In this study, the performance of two Extreme Learning Machine (ELM) and Support Vector Classification (SVC) base models was cross-validated using a compiled dataset. To further enhance predictive models, optimization techniques like the Mother Optimization Algorithm (MOA) and Stochastic Paint Optimizer are used. Also, feature importance analysis and the Wilcoxon test were utilized to evaluate the significance of individual features and to statistically compare the performance of the models, confirming that the improvements achieved by the optimized models were statistically significant. Results indicate that these optimizations also greatly enhanced model performance. For instance, ELMA (MOA-optimized ELM) outperformed the baseline ELM model by boosting test accuracy from 91.29 to 94.52% and precision from 91.84 to 94.56%. Similarly, SVCA (MOA-optimized SVC) achieved a 93.55% test accuracy, higher than the baseline SVC's 90.00%. Overall, ELMA delivered the best performance, with precision scores of 0.94 and 0.92 in mismatched and matched cases, while the base SVC showed the weakest results, with 0.98 and 0.70 precision in those conditions. Integrating optimization algorithms with ML models substantially boosts the accuracy of predicting successful matches in speed dating, offering a promising approach to improving dating outcomes.