No abstract

No abstract

Process mining provides process improvement in a variety of application domains. A primary focus of process mining is transferring information from event logs into process model. One of the issues of process mining is dealing with invisible prime tasks. An invisible prime task is an additional task in the process model to assist in showing real processes. However, a few of algorithm solves the issue. This research proposes an algorithm for dealing with invisible prime tasks. The proposed algorithm contains rules and equations utilizing probability of state transition of Coupled Hidden Markov and double time-stamped in event logs. The rules and equations are used for determining invisible prime tasks and parallel control-flows patterns. In addition to dealing with invisible prime tasks, the experiment results also show that the proposed algorithm obtains right parallel control-flow patterns from non-complete event logs. This proposed algorithm also decreases usage of the invisible prime task in A# algorithm without reducing the quality of discovered process models. It has proven with the fitness of process models obtained by the proposed algorithm are relatively high as those obtained by A# algorithm.

Existing methods, such as Graph Edit Distance (GED) and Cosine measure, still have drawback in obtaining similarity of parallel relationships by neglecting the control-flow patterns, i.e. AND, OR, and XOR. Since AND > OR > XOR, the similarity value of AND versus OR is greater than XOR versus OR and AND versus XOR. This paper proposes two new similarity methods, Tree Declarative Pattern Edit Distance (TPED) and Cosine-Tree Declarative Pattern (Cosine-TDP). They provides value to the control-flow pattern so the value of similarity can be seen more differently. The new methods utilize tree model of the declarative pattern. The results show that the proposed methods are better at differentiating parallel relationships than the existing methods, GED and Cosine measure. In obtaining AND versus OR, XOR versus OR, and AND versus XOR, TPED obtained 0.821, 0.811, and 0.78 while Consine-TDP obtained 0.834, 0.826, and 0.693. Meanwhile, GED obtained 1 for all parallel relationships whereas Cosine measure obtained 0.02, 0.08, and 0.04.

No abstract

An event log records the business processes of a company. Modeling event logs aim to help users in analyzing business processes. One of the problems in modeling event logs automatically is the addition of invisible tasks. Invisible tasks are dummy activities, other than activities of an event log, that are added to a process model to describe a correct process model. This research proposes a graph-based algorithm to mine the data from an event log. From the data, the graph-based algorithm establishes an additional-invisible-task process model by converting all of the processes in the event log into a link list and adding invisible tasks and operators for parallel relations, such as XOR Split or XOR Join. The experimental analysis explains that the fitness of the discovered process models by the graph-based algorithm was as high as that of compared algorithms, such as Alpha# models, Alpha$ models, CHMM-NCIT models, and CHMM-IT models. Furthermore, the graph-based algorithm is more efficient than existing algorithms. This was proven by the time complexity of the graph-based, which is O(n 2 ) while both of Alpha# and Alpha$ algorithm have a time complexity of O(n 4 ) and both of CHMM-IT and CHMM-NCIT algorithm have O(n 3 ).

Many restaurant review analysis have been done, however only few analysis have been done for specific aspects of a restaurant. In this context this paper proposes aspect based restaurant analysis which includes Physical environment, Food quality, Service quality and Price fairness. The analysis steps include Aspect Term Extraction (ATE), Aspect Keyword Extraction (AKE), Aspect Categorization (AC) and Sentiment Analysis (SA). ATE employs the modification of Double Propagation method and several Topic Modelling methods, AKE utilizes Term Frequency-Inverse Cluster Frequency (TF-ICF), in AC we propose Hybrid ELMo-Wikipedia (HEW), and in SA we propose Hybrid Expanded Opinion Lexicon-SentiCircle (HEOLS). The results show that our modification of the methods used in ATE could increase the f1measure of the AC by average 2%, then the HEW that we proposed had better f1measure compared to other similar methods by average 6%. Other than that, our proposed HEOLS can expand and redetermine the Opinion Lexicon polarity and can increase f1measure of SA by 6%.

Many corporations worldwide use an enterprise resource planning (ERP) system to manage their business process, which continuously changes due to dynamic business requirements [1]. Because the processes run continuously, ERP produces a considerable log of processes. Manual observation will have difficulty monitoring the sizeable log, especially detecting anomalies. It needs the method that can detect anomalies in the huge log. Standard business processes are usually incorporated into standard operating procedures (SOP), which are used as a reference to find any deviations. Deviations or anomalies in the business process can be caused by variations or operation errors [2]; however, some of the anomalies may be the result of fraudulent behaviours [3]. Fraud can be committed in many ways and can lead to significant losses. In 2012, the Association of Certified Fraud Examiners (ACFE) reported that there had been 1.388 fraud cases in 96 countries, which have incurred US$1.4 billion in losses [4]. On average, organizations