I need a method based on Low process and memory for Real Time apllication
today we used many algorithm such as GA, Neural Network and Aprior Algorithm, The above methods has a high complexity, processing time and memory space requirements.
you can use apriori algorithm with big data data frameworks shch as spark or flink
you can use apriori algorithm with big data data frameworks shch as spark or flink
Dear Mohammad,
This can be can be done based on several algorithms. One of them is the Brute-Force algorithm that identifies all possible ARs. Then, it prunes groups that do not reach the predefined Support threshold value. However, this algorithm is computationally prohibitive since that the identification of all groups, corresponding to N classes, needs 2^N time complexity. Another algorithm is the apriori algorithm that utilizes the property of anti-monotone, which means that if a AR is considered as infrequent, then all of its supersets must be infrequent as well. Thus, they do not need to be generated. However, this algorithm still has to generate the candidate ARs. For instance, suppose that we have 104 frequent items of of size 1, it requires to generate about 107 groups (candidate ARs) of size 2. Furthermore, it needs to generate about 1030 candidate ARs of size 10. Thus, this algorithm does not work in the situation where low Support threshold values are selected. Another algorithm is the Frequent-Pattern Growth (FP Growth) algorithm. In this algorithm, there is no need to produce the candidate ARs. Instead, it uses a divide-and-conquer technique to mine FUPs. It firstly builds a special data structure called Frequent-Pattern tree (FP-tree). This tree is used to compress information of class associations. Then, FP Growth divides the FP-tree into a collection of databases, such that each one is related to one frequent group of classes.
Among the presented algorithms, FP Growth is the best one since that it outperforms the others in terms of time and space complexity.
Best,
- Badges
- Science topic