基于情境感知和序列模式挖掘的气象学习资源推荐算法

随着互联网的快速发展，气象部门职工作为学习者可以获得的学习资源得到极大丰富。信息超载导致检索合适的在线学习资源时遇到了困难；学习者在不同学习环境和序列访问模式上也有不同的学习需求。但是，现有的推荐系统，如基于内容的推荐和协同过滤，没有结合学习者的情境和序列访问模式，推荐结果准确度不高。本文提出了一种结合情境感知、序列模式挖掘和协同过滤算法的混合推荐算法来为学习者推荐学习资源。混合推荐算法中，情境感知被用来整合学习者的情境信息，如知识水平和学习目标；序列模式挖掘被用来对网络日志进行挖掘，发现学习者的序列访问模式；协同过滤被用来根据学习者的情境数据和序列访问模式为目标学习者计算预测并生成建议。实验和应用效果表明，该混合推荐算法推荐的质量和准确性方面优于其他推荐算法。

Meteorology Learning Resource Recommendation Algorithm Based on Context Awareness and Sequential Pattern Mining

With the rapid development of the Internet, the learning resources available to meteorological staff as learners are greatly enriched. Information overload leads to difficulties in retrieving suitable online learning resources; learners also have different learning needs in different environments and sequential access modes. However, existing recommendation systems, such as collaborative filtering and content-based recommendation, only involve two types of entities: items and users. They do not consider contextual information such as learners’ learning objectives and knowledge levels, as well as different sequential access patterns to learning resources, resulting in low accuracy in recommendation results. This paper proposes a hybrid recommendation algorithm that combines context awareness, sequential pattern mining, and collaborative filtering algorithms to recommend learning resources for learners. The hybrid recommendation algorithm includes three main steps: (1) integrating contextual information into the recommendation process using a contextual pre-filtering algorithm, (2) calculating learner similarity based on contextualised data and predicting the evaluation of learning resources, (3) generating the first N recommendations for the target learner, applying the GSP algorithm to the results, and filtering the final recommendations based on the learner’s sequential access patterns. In hybrid recommendation algorithms, context awareness is used to integrate contextual information about learners, such as knowledge level and learning objectives; sequential pattern mining is used to mine weblogs to discover learners’ sequential access patterns; collaborative filtering is used to calculate predictions and generate recommendations for targeted learners based on contextual data and sequential access patterns of learners. This hybrid recommendation algorithm incorporates contextual characteristics and learners’ sequential access patterns into the recommendation process to achieve improved personalised recommendation. When calculating the similarity between learners and learning items, the contextual characteristics of learners are taken into account; combining multiple recommendation techniques helps alleviate data sparsity problems. Experimental comparisons have shown that this recommendation algorithm is significantly superior to other recommendation algorithms in terms of recall, accuracy, and F1, especially when the neighbourhood value is 25. The hybrid recommendation algorithm is applied to the Yunzhipei intelligent teaching management system, with a user satisfaction rate of 93.7%, achieving a good application effect. In later stages, hybrid recommendation algorithms will be applied to the search and recommendation of electronic documents and institutional trees, providing assistance to meteorological employees in recommending accurate reference documents; it can also be combined with ElasticSearch to re locate and value mine heterogeneous data, enhancing the value of business and management historical data.