Event‐Tree Based Sequence Mining Using LSTM Deep‐Learning Model

Abstract

During the operation of modern technical systems, the use of the LSTM model for the prediction of process variable values and system states is commonly widespread. The goal of this paper is to expand the application of the LSTM-based models upon obtaining information based on prediction. In this method, by predicting transition probabilities, the output layer is interpreted as a probability model by creating a prediction tree of sequences instead of just a single sequence. By further analyzing the prediction tree, we can take risk considerations into account, extract more complex prediction, and analyze what event trees are yielded from different input sequences, that is, with a given state or input sequence, the upcoming events and the probability of their occurrence are considered. In the case of online application, by utilizing a series of input events and the probability trees, it is possible to predetermine subsequent event sequences. The applicability and performance of the approach are demonstrated via a dataset in which the occurrence of events is predetermined, and further datasets are generated with a higher-order decision tree-based model. The case studies simply and effectively validate the performance of the created tool as the structure of the generated tree, and the determined probabilities reflect the original dataset.

Prediction of Complex Event Graphs with Neural Networks

“…Abonyi et al [7] expanded the seq2seq algorithm so that its output is a probability tree, that describes the predicted alternative courses of events, instead of a single, most probable sequence. As trees do not allow for modeling complicated structures, like parallelism and conditional forking, our NN-based approach aims to produce event graphs that contain complex control flow elements as well.…”

“…The studies in recent years show that beam search techniques usually provide much better accuracy [6]. The beam search method is used in [7] to generate a probability tree that describes the network of alternative events based on a given input (see Figure 1). Figure 1.…”

“…Figure 1. Sample probability event tree [7] The generated probability event tree can provide more information than conventional event sequences, but real event graphs may have much more complex structures. In many application areas, the event graph pattern schema may contain loops or parallel sequences of different actors.…”

Network-based visualisation of frequent sequences

“…Frequent sequence pattern mining is widely used to extract knowledge from event log files, e.g., it has been successfully applied for user activity pattern analysis [1], workload prediction [2], malware behaviour analysis [3], and prediction of transition probability [4]. Frequent sequence pattern mining algorithms do not provide sufficient information on the relationship and context of the large number of sequences extracted, making the interpretation and utilisation of the results difficult.…”