摘要
针对传统机器学习方法构建多元化业务模型中存在分析大量数据困难,在业务数据搜索中缺乏有效整合人类经验和偏好的能力的问题。提出一种通过整合机器学习和交互式遗传算法达成交互式学习的新型机器学习方法。交互式学习能够通过各种人机交互工具和数据将领域专家直接纳入模型构建过程中;使用交互式学习,用户可以通过选择参数,评估和比较模型,选择最适合的模型来训练学习算法。实验表明,交互式学习算法收敛所需的训练次数更少,易达到最优解。
Based on the problems of the traditional machine learning methods in building a diversified business model,there are several problems,including difficulty in analyzing large amounts of data,the lack of the ability to effectively integrate human experience and preferences in business data search.A new type of machine learning method for interactive learning and interactive genetic algorithm is proposed.Interactive learning can directly incorporate domain experts into the model building process through various human-computer interaction tools and data.By using this method,users could select a model that is most suitable to their goals to train learning algorithms through selecting parameters,and then evaluate and compare model.The results show that the interactive learning algorithm requires less training times to converge,and it is easy to reach the optimal solution.
作者
程哲
胡超
辛阔
唐卓尧
吴小刚
CHENG Zhe;HU Chao;XIN Kuo;TANG Zhuo-yao;WU Xiao-gang(China Southern Power Grid Power Dispatching Control Center,Guangzhou 510663,China;NARI Information&Communication Technology Co.,Ltd.,Nanjing 210003,China)
出处
《信息技术》
2021年第12期49-54,共6页
Information Technology
基金
中国南方电网有限责任公司科技项目(ZDKJXM201-80074)。
关键词
机器学习
交互式学习
遗传算法
主观评估值
权重
machine learning
interactive learning
genetic algorithm
subjective evaluation value
weight
