Complex Sequential Data Analysis: A Systematic Literature Review of Existing Algorithms

Abstract

This paper provides a review of past approaches to the use of deep-learning frameworks for the analysis of discrete irregularpatterned complex sequential datasets. A typical example of such a dataset is financial data where specific events trigger sudden irregular changes in the sequence of the data. Traditional deep-learning methods perform poorly or even fail when trying to analyse these datasets. The results of a systematic literature review reveal the dominance of frameworks based on recurrent neural networks. The performance of deep-learning frameworks was found to be evaluated mainly using mean absolute error and root mean square error accuracy metrics. Underlying challenges that were identified are: lack of performance robustness, non-transparency of the methodology, internal and external architectural design and configuration issues. These challenges provide an opportunity to improve the framework for complex irregular-patterned sequential datasets.

Selecting Datasets for Evaluating an Enhanced Deep Learning Framework

“…Step 1-The identification of existing sequential datasets, analysis artefacts, and evaluation criteria. This step is an extension of the systematic literature review work [4] which reviewed over 400 research articles from year 2015 to 2020 and narrowed them to the 33 most relevant articles. A summary of identified sequential models, architecture, datasets and evaluation techniques is shown in Table I.…”

“…There is still not much agreement on how to solve the challenges posed by the analysis of irregular sequential datasets [4]. Contrary to existing frameworks-which are skewed towards probabilistic randomization or ad-hoc design approaches, which are prone to accuracy, stability, explainability and repeatability deficiencies-the SeLFISA framework aims at addressing these deficiencies [22].…”

“…The underlying challenges associated with existing deep learning frameworks for the analysis of discrete irregular patterned complex sequences were identified [4]. Specific concerns raised were: performance robustness; transparency of the methodology; literature consistency; internal and external architectural design and configuration issues.…”

“…1, a class of datasets with irregular patterns were identified as a key stage of developing better deep learning artefacts for sequential analysis. Seventy three sequential datasets from eight domains of different sources were extracted from 400 research articles, by means of a systematic literature review process [4]. Datasets, characterised with the most irregular patterns, were then identified for the evaluation of the deep learning framework.…”