High Utility Itemset Mining Research Articles

Mining high utility itemsets with time‐aware scheduling using Apache Spark

SummarySince the last decade, Market Basket Analysis has been propelled by augmentation of revenue information. Termed as high utility itemset mining (HUIM), this task considers the factors of purchase quantity and unit profit of the items in the transaction database. Although several sequential algorithms to mine HUIs exist, their performance degrades as the database becomes voluminous. Distributed computing solutions such as Apache Hadoop and Apache Spark have proven effective in alleviating this bottleneck. In this regard, the current study develops a parallel workflow to adapt a single‐phase tree‐based algorithm called the single phase utility computation (SPUC) algorithm on a Spark cluster. Based on the time taken to mine individual conditional pattern bases in SPUC, an assignment strategy that partitions the search space across the cluster is proposed in parallel SPUC (PSPUC) algorithm. Experimental evaluation conducted using real and synthetic datasets demonstrate that PSPUC outperforms PHUI‐Growth algorithm. Apart from this, PSPUC in conjunction with the time‐aware assignment strategy converges mining faster than a random assignment of items. A linear speedup of PSPUC is also demonstrated.

An efficient utility-list based high-utility itemset mining algorithm

An efficient utility-list based high-utility itemset mining algorithm

High-utility itemsets mining based on binary particle swarm optimization with multiple adjustment strategies

As an essential data mining task, high-utility itemset mining (HUIM) has attracted a lot of research. With the increase of dataset size, traditional exact HUIM algorithms are faced with exponential growth search space, which is unacceptable for some algorithms. As HUIM can be treated as a combinatorial optimization problem, Evolutionary Computation (EC) based HUIM approaches have been proposed and shown promise performance in mining HUIs. However, the existing EC-based HUIM approaches usually only find part of the HUIs in a limitation time or discovering all the HUIs is usually time-consuming. In this study, an improved binary particle swarm optimization for HUIM (HUIM-IBPSO) is proposed with multiple adjustment strategies to address these problems. In HUIM-IBPSO, a particle movement direction adjustment strategy is presented to keep the same HUIs during the evolution process. In order to utilize the repeated HUIs more efficiently and enhance the search ability, the strategy of local exploration is proposed in HUIM-IBPSO. A restart strategy for the population is developed in HUIM-IBPSO with the purpose to avoid the premature convergence before discovering any HUIs. To mine HUIs more efficiently, particle modify strategy and fitness value hash strategy are introduced in HUIM-IBPSO. A comprehensive comparison with five state-of-the-art EC-based HUIM algorithms and three precise HUIM algorithms on real datasets shows that the designed model outperforms them in terms of the number of HUIs found, the speed at which they converge, and the duration of execution (runtime).

H-Map-Based Technique for Mining High Average Utility Itemset

High Utility Itemset Mining (HUIM) is the process of locating itemsets that are profitable and useful to users. One of the key flaws in HUIM is that as the length of the itemset increases, the utility also increases. The true utility/profit of the itemset is not revealed in HUIM. High Average Utility Itemset mining overcomes the limitations of HUIM by taking the length of the itemset into account when estimating the utility. Existing pruning methods used for eliminating weak candidates overestimate the average usefulness of itemsets, causing the mining process to slow down. To prevent processing unpromising candidate itemsets and efficiently reduce the search space and processing time, the proposed methodology employs Upper Bound using Remaining Items Utility, Maximum Itemset Utility, and Sum of Maximum Utility in a Transaction. It also uses the multithreaded parallel approach to reduce the processing time. The H-Map-based data structure (H-Map) used for storing the utility values reduces the lookup time and joins used for itemset extension compared to existing state-of-the-art High Average Utility Itemset mining algorithms. The performance of the proposed work is evaluated in terms of memory usage and the time taken for processing. The proposed work increases the overall efficiency of the system employing effective pruning algorithms for pruning poor candidate itemsets and an efficient data structure for storing utility values.

Design and Implementation of Automatic Selection of the Most Efficient Itemset Algorithm Based on Spark

The combination of Spark distributed platform and High-Utility Itemset Mining can solve the problem of long running time issue of High-Utility Itemset Mining. In the experiment, we conclude that Spark-based parallel D2HUP and EFIM algorithms have a greater improvement in running time efficiency than serial algorithms. The existing research has shown that the EFIM and D2HUP algorithms are the two best algorithms for High-Utility Itemset Mining. This paper generates 118 datasets by generating and collecting the running time of the two algorithms in the real and simulated datasets, taking into account the characteristics of each dataset’s length, sparse degree, and dataset size as characteristics with running time as the prediction target and then establishing a model. The accuracy of the prediction was evaluated through experiments, and a set of rules based on decision trees was generated. According to the rules, the fastest algorithm between EFIM and D2HUP can be predicted very well.

PUC: parallel mining of high-utility itemsets with load balancing on spark

Abstract Distributed programming paradigms such as MapReduce and Spark have alleviated sequential bottleneck while mining of massive transaction databases. Of significant importance is mining High Utility Itemset (HUI) that incorporates the revenue of the items purchased in a transaction. Although a few algorithms to mine HUIs in the distributed environment exist, workload skew and data transfer overhead due to shuffling operations remain major issues. In the current study, Parallel Utility Computation (PUC) algorithm has been proposed with novel grouping and load balancing strategies for an efficient mining of HUIs in a distributed environment. To group the items, Transaction Weighted Utility (TWU) values as a degree of transaction similarity is employed. Subsequently, these groups are assigned to the nodes across the cluster by taking into account the mining load due to the items in the group. Experimental evaluation on real and synthetic datasets demonstrate that PUC with TWU grouping in conjunction with load balancing converges mining faster. Due to reduced data transfer, and load balancing-based assignment strategy, PUC outperforms different grouping strategies and random assignment of groups across the cluster. Also, PUC is shown to be faster than PHUI-Growth algorithm with a promising speedup.

An Enhanced Fast – High Utility Item set Mining Method for Large Datasets

High Utility Itemset mining is considered one of the critical and challenging problems in data mining. The existing mining framework is limited to analyzing occurrence counts of items in the Database. However, this framework applies a single minimum utility threshold value that fails to consider different item characteristics. Recent methods of association mining focused on finding the high utility itemsets instead of frequent itemsets generations. Some utility-based mining methods that is Faster High Utility Itemset Mining (FHM), High Utility Itemset Miner (HUI-Miner), Direct Discovery of High Utility Patterns (D2HUP), Utility Pattern Growth (UP Growth & UP Growth+) are studied for the generation of high utility itemsets generations. Existing HUI mining methods are effectively generating HUIs. However, developing a faster and memory-efficient HUI mining method is required. For this purpose, this work develops an Enhanced Fast - High Utility Itemset Mining (EF-HUIM) method for the faster generation of high utility itemsets and respective association rules.

High Utility Item-Set Mining From Retail Market Data Stream With Various Discount Strategies

High Utility Item-set Mining (HUIM) is the futuristic remodel version of Frequent Item Mining (FIM). It has the ability to discover customer purchase trends in the retail market. By using that knowledge, retailers can incorporate innovative schemes (discounts, cross-marketing, seasonal sales offers,... etc) to enhance profit. Even though many HUIM algorithms are proposed to detect profitable patterns, most of them cannot be applied to all kinds of retail market data sets due to certain assumptions. The first assumption is that the items always produce a positive profit. But in reality, even though overall profit could be positive, some of the items make negative profit. The second one is they are developed for static transactional data. Those are useful to take decisions at some intervals like quarterly, half-yearly, yearly. But, to take decisions at any time by analyzing the present sales trend, it is required to process the data stream.

UBP-Miner: An efficient bit based high utility itemset mining algorithm

HUIM (High utility itemset mining) is a key problem in data mining. The goal is to find itemsets having a high importance or profit in a database, to identify useful knowledge that can support decision-making. In recent years, many HUIM algorithms have been put forward. Among them, utility-list-based algorithms have become very popular as they are easily extendable and efficient. Although several improvements were made, efficiency remains a critical issue. To address this problem, this paper proposes to improve the utility-list construction process, a key operation that has not been much studied in prior work. A novel set of bitwise operations is proposed called BEO (Bit mErge cOnstruction) to speed up the construction process. Besides, a novel data structure called UBP (Utility Bit Partition) is designed to support BEO. This structure is integrated into a novel UBP-Miner algorithm, which also applies several search space reduction strategies. Experimental results show that UBP-Miner is faster than several state-of-the-art algorithms such as HUI-Miner* and ULB-Miner on common benchmark datasets.

High Median Utility Itemset Mining for recovering streaming window transaction using novel Modified Heap‐based Optimization

AbstractHigh utility itemsets (HUIs) are items in the dynamically streaming transaction list that generate a high‐profit margin. Many of the real‐time applications depend on finding HUIs from the transaction list. However, HUI mining is time‐consuming and results in high complexity due to memory requirements, a large search space, and the cost of HUI estimation. To overcome these issues we have proposed a novel High Median Utility Itemset Mining (HMUIM) approach. This approach utilizes the HMUI‐Miner which effectively ignores the unnecessary itemsets, that is, items with less profit, and mines the HUIs from the database. Furthermore, it has the limitation of deleting the transaction list based on the sliding window size used while conducting a dynamic streaming dataset. To tackle this issue we proposed a novel Modified Heap‐based Optimizer (MHBO) algorithm which effectively copies the HUIs transaction list and preserves it for further process. The MHBO is the combination of Heap Based Optimizer and nine fuzzy rules. The fuzzy rules are used to analyze the priority of the transaction list and based on that it ignores the transaction list with low priorities. Experimental analysis is performed for the proposed method on the real‐time dataset and compared with HUI and HAUIM approaches. The proposed method reduces the execution time and memory usage based on the minimum threshold. Meanwhile, the proposed MHBO approach is compared with iMEFIM, REX, and FCHUIM state‐of‐art works based on the sliding window concept. The execution time to store the replicated copies of original data is less for our proposed method.

Performance comparison of inertia weight and acceleration coefficients of BPSO in the context of high-utility itemset mining

Performance comparison of inertia weight and acceleration coefficients of BPSO in the context of high-utility itemset mining

A survey on soft computing-based high-utility itemsets mining

A survey on soft computing-based high-utility itemsets mining

Efficient evolutionary computation model of closed high-utility itemset mining

Efficient evolutionary computation model of closed high-utility itemset mining

Compact Tree Structures for Mining High Utility Itemsets

High Utility Item set Mining (HUIM) from large transaction databases has garnered significant attention as it accounts for the revenue of the items purchased in a transaction. Existing tree-based HUIM algorithms discard unpromising items and require at most two database scans for their construction. Hence, whenever utility threshold is changed, the trees have to be reconstructed from scratch. In this regard, the current study proposes to not only incorporate all the items in the tree structure but compactly represent transaction information. The proposed trees namely- Utility Prime Tree (UPT), Prime Cantor Function Tree (PCFT), and String based Utility Prime Tree (SUPT) store transaction-level information in a node unlike item-based prefix trees. Experiments conducted on both real and synthetic datasets compare the execution time and memory of these tree structures with a proposed Utility Count Tree (UCT) and existing IHUP, UP-Growth trees. Due to transaction-level encoding, these structures consume significantly less memory when compared to the tree structures in the literature.

Generic Itemset Mining Based on Reinforcement Learning

One of the biggest problems in itemset mining is the requirement of developing a data structure or algorithm, every time a user wants to extract a different type of itemsets. To overcome this, we propose a method, called <i>Generic Itemset Mining based on Reinforcement Learning</i> (GIM-RL), that offers a unified framework to train an agent for extracting any type of itemsets. In GIM-RL, the environment formulates iterative steps of extracting a target type of itemsets from a dataset. At each step, an agent performs an action to add or remove an item to or from the current itemset, and then obtains from the environment a reward that represents how relevant the itemset resulting from the action is to the target type. Through numerous trial-and-error steps where various rewards are obtained by diverse actions, the agent is trained to maximise cumulative rewards so that it acquires the optimal action policy for forming as many itemsets of the target type as possible. In this framework, an agent for extracting any type of itemsets can be trained as long as a reward suitable for the type can be defined. The extensive experiments on mining high utility itemsets, frequent itemsets and association rules show the general effectiveness and one remarkable potential (agent transfer) of GIM-RL. We hope that GIM-RL opens a new research direction towards learning-based itemset mining.

Implementing reverse up growth tracking approach under distributed data mining

Data mining is the methodology which discovers useful and hidden information from large databases. Many researchers have proposed innumerable algorithms in the field of data mining. In this system improvised UP-growth is considered for mining high utility itemset from potential high utility itemset and improvised under different constraints. The node utility and reorganised transaction utility are the aspects considered as the key term in the proposed system which are manipulated using the technique as in UP-growth. However, mining potential high utility itemset from RTU using UP-growth needs number of tree traversals. This is reduced in the proposed system by introducing bottom-up approach and merging certain manipulations. However, working the system as a sequential process will be time consuming. Distributed environment is considered in the proposed system to overwhelm the problem in existing methodology.

EAHUIM: Enhanced Absolute High Utility Itemset Miner for Big Data

High utility itemset mining (HUIM) is a data mining technique that identifies the itemsets with utility levels exceeding a pre-determined threshold. The factor utility is described as the combination of magnitude and element of significance for an item, and the algorithm objectives to locate the set of items with a utility higher or equivalent to a set benchmark. These itemsets are utilized to build association rules for data mining systems. However, in the age of big data, conventional (HUIM) strategies are least effective with limited processing capabilities. This work proposes an optimized technique, Enhanced Absolute High Utility Itemset Miner (EAHUIM) by incorporating various refinements into the Absolute High Utility Itemset Miner (AHUIM) Algorithm. EAHUIM discovers the itemsets from large datasets in near real-time and serves as the foundation for information management and decision-making systems by providing diverse insights. The experimental analysis reveals that EAHUIM outclasses other state-of-the-art algorithms for HUIM.

TKN: An efficient approach for discovering top-k high utility itemsets with positive or negative profits

Top-k high utility itemsets (HUIs) mining permits discovering the required number of patterns - k, without having an optimal minimum utility threshold (i.e., minimum profit). Multiple top-k HUIs mining algorithms have been introduced with interesting results. However, these algorithms focus mainly on mining patterns from positive profit datasets, while few preliminary studies can handle datasets with negative profits. Moreover, conventional top-k HUI mining algorithms, that are meant for exploring positive profit datasets, perform poorly when mining top-k HUIs on highly dense and large datasets. In this paper, we propose TKN (efficiently mining Top-K HUIs with positive or Negative profits) which employs generalized and adaptive techniques to mine both positive and negative profit datasets effectively. The proposed approach adopts transactional projection and merging mechanisms to decrease the dataset traversing cost. Furthermore, several pruning and threshold elevating ideas are utilized to significantly narrow the exploration space. To highlight the reliability of the devised TKN, a series of extensive comparisons were conducted using two versions of six real datasets. The obtained results reveal that TKN is clearly superior in finding the required number of patterns, whether on positive or negative profit datasets, compared to the current cutting-edge competitors.

Efficient mining of cross-level high-utility itemsets in taxonomy quantitative databases

In contrast to frequent itemset mining (FIM) algorithms that focus on identifying itemsets with high occurrence frequency, high-utility itemset mining algorithms can reveal the most profitable sets of items in transaction databases. Several algorithms were proposed to perform the task efficiently. Nevertheless, most of them ignore the item categorizations. This useful information is provided in many real-world transaction databases. Previous works, such as CLH-Miner and ML-HUI Miner were proposed to solve this limitation to discover cross-level and multi-level HUIs. However, the CLH-Miner has a long runtime and high memory usage. To address these drawbacks, this study extends tight upper bounds to propose effective pruning strategies. A novel algorithm named FEACP (Fast and Efficient Algorithm for Cross-level high-utility Pattern mining) is introduced, which adopts the proposed strategies to efficiently identify cross-level HUIs in taxonomy-based databases. It can be seen from a thorough performance evaluation that FEACP can identify useful itemsets of different abstraction levels in transaction databases with high efficiency, that is up to 8 times faster than the state-of-the-art algorithm on the tested sparse databases and up to 177 times on the tested dense databases. FEACP reduces memory usage by up to half over the CLH-Miner algorithm.

An efficient HUIBA Algorithm for Mining High Utility Itemset

Utility Mining is to spot the itemsets with highest utilities, by considering profit, quantity, cost or other user preferences. Mining High Utility itemsets from a transaction database is to seek out itemsets that have utility above a user-specified threshold. Bio inspired algorithm is extremely efficient for mining High Utility Itemset(HUI), but it will not find all HUI in the database and the quality is poor within the number of discovered HUI. A replacement framework using BA algorithm is proposed to rectify this issue. The proposed algorithm is more efficient in terms of quality and convergence speed when put next to other algorithms.