English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

In online educational courses, knowledge point is the basic unit used for the organization and dissemination of information. Mining the association of knowledge points can help to improve the navigation of course learning. However, it is difficult for the existing mining algorithms to balance the performance and accuracy of mining the association of knowledge points. This paper proposes a FP-tree based algorithm used for mining the association of knowledge points. With the multiple minimum support strategy, this algorithm modifies the setting mode of each item's minimum support and improves the construction method of FP-tree to adapt to the dynamic database mining. The experiment shows that the algorithm improved the accuracy of the mining results and ensured a good performance. Introduction According to the global education industry report from GSV (Global Silicon Valley), the market size of global online education is expected to increase to $ 255.5 billion in 2017. More and more people start to learn knowledge by online education. Knowledge point (defined in reference [3]) is the basic unit used for the organization and dissemination of information in online education. Mining the association of knowledge points is of great significance for improving the navigation and recommendation of courses. Nowadays most researches about the association of knowledge points use traditional data mining algorithms to analyze the users’ behavior data directly, without improvement and optimization combined with the features of the association of knowledge points. Hence, the performance and accuracy can’t be obtained at the same time in these algorithms. Therefore, it is necessary to find a specialized algorithm used for mining the sets of frequent knowledge points. In this paper, by combining with the features of the association of knowledge points, a new algorithm will be proposed. This algorithm simplifies the setting mode of each item's minimum support in the multiple minimum support strategy and improves the construction method of FP-tree to adapt to the dynamic database mining. Furthermore, a new data structure named frequent conditions pattern filter list is introduced into the algorithm, with the help of the list, the algorithm will have a higher performance in the situation where the minimum support changes continually. In the experiment, original data collected from SLMS (Sophia Learning Manage System) is 70 students’ online learning record data of the Android App Development Course. By comparison of the results, the algorithm is better adapted to mining the association of knowledge points. Related work In the research area of mining the association of knowledge points, most researches process and analyze the users’ behavior data to get the association of knowledge points by traditional data mining algorithms. Therefore, with the improvement and innovation of association rule mining algorithm, the research for association of knowledge points also progresses constantly. Liu. B et al. proposed MSapriori algorithm which allows user to specify multiple minimum item supports. This algorithm solved rare item problem, but the performance of the algorithm wasn’t enhanced obviously. J. Han et al. proposed FP-growth algorithm which uses FP-tree to find the set of frequent pattern, and FP-tree is 6th International Conference on Information Engineering for Mechanics and Materials (ICIMM 2016) © 2016. The authors Published by Atlantis Press 34 an extended prefix-tree for storing compressed and crucial information about frequent patterns. FP-growth algorithm is much more efficient than Apriori algorithm. Hu, Y. et al. proposed an extended FP-growth algorithm combined with multiple minimum support model. However, each minimum support of item need to be specified by user. Yang, H. et al. proposed a multi-support mining algorithm based on knowledge points. In this algorithm, each minimum support of item will be calculated from the two aspects which include learning and testing for the corresponding knowledge point. But the calculation method of minimum support is too complex and the algorithm is not very efficient. A FP-Tree-Based Algorithm for mining the sets of frequent knowledge points. Compression of transaction database. FP-tree (defined in reference [2]) is a specific data structure used for storing the compressed frequent patterns in transaction database. After building the FP-tree for the transaction database, all of the frequent pattern can be obtained from FP-tree by FP-growth algorithm. But FP-tree only stores the information of frequent item. Thus FP-growth algorithm is only used for the static database. If the transaction database changes, the FP-tree needs to be built again. This paper presents an improved FP-tree data structure named DFP-tree (Dynamic Frequent Pattern tree) which can be better adapted to the dynamic database mining. Instead of only storing the information of frequent item, all the transaction database will be compressed and stored into DFP-tree during the building process of DFP-tree. According to Fig. 1, this paper introduce the building process of DFP-tree through an example. Algorithm 1(DFP-tree construction) Input: transaction database DB. Output: Its DFP-tree. Method: 1. For each transaction in DB to get the actual support of each item in DB. 2. Create the root of a DFP-tree T, and label it as “null”. 3. For each transaction t in DB do the following: (1). Sort all items in t according to their actual support in decreasing order. (2). Call insert-tree([p|P],T). Algorithm 2(DFP-tree dynamic data processing) Input: newly added transaction data NDB, the DFP-tree of previous DB. Output: updated DFP-tree Method: 1. For each transaction in NDB to get the actual support of each item in NDB. 2. Update the support of each item and sort according to their actual support in decreasing order. 3. Check whether the index of each item in DB changed. If all of the indexes didn’t change: For each transaction t in NDB do the following. (1). Sort all items in t according to their actual support in decreasing order. (2). Call insert-tree ([p|P], T). Else call Algorithm 1 to rebuild DFP-tree. Function insert-tree([p|P], T) (p is the first element and P is the rest in sorted items [p|P].) While (P is not null){ If T has a child N whose item-name = p.item-name then N.count++; else create a new node N, and N.count = 1; let N.parent-link link to T; let N.node-link link to the node which has the same item-name. } Fig. 1 The algorithms for DFP-tree construction.

A FP-Tree Based Algorithm for mining knowledge point association