Multi-view contrastive self-supervised learning of accounting data representations for downstream audit tasks

Abstract

International audit standards require the direct assessment of a financial statement's underlying accounting transactions, referred to as journal entries. Recently, driven by the advances in artificial intelligence, deep learning inspired audit techniques have emerged in the field of auditing vast quantities of journal entry data. Nowadays, the majority of such methods rely on a set of specialized models, each trained for a particular audit task. At the same time, when conducting a financial statement audit, audit teams are confronted with (i) challenging time-budget constraints, (ii) extensive documentation obligations, and (iii) strict model interpretability requirements. As a result, auditors prefer to harness only a single preferably 'multi-purpose' model throughout an audit engagement. We propose a contrastive self-supervised learning framework designed to learn audit task invariant accounting data representations to meet this requirement. The framework encompasses deliberate interacting data augmentation policies that utilize the attribute characteristics of journal entry data. We evaluate the framework on two real-world datasets of city payments and transfer the learned representations to three downstream audit tasks: anomaly detection, audit sampling, and audit documentation. Our experimental results provide empirical evidence that the proposed framework offers the ability to increase the efficiency of audits by learning rich and interpretable 'multi-task' representations.

Symmetrical and Asymmetrical Sampling Audit Evidence Using a Naive Bayes Classifier

“…Schreyer et al [8,9] constructed an autoencoder neural network to sample journal entries in their two papers. They fed attributes of these journal entries into the resulting autoencoder.…”

Federated and Privacy-Preserving Learning of Accounting Data in Financial Statement Audits

“…Nowadays, applied representation learning techniques encompass, autoencoder neural networks [53], adversarial autoencoders [51], or variational autoencoders [68]. Lately, self-supervised learning techniques have been proposed to learn rich representations for multiple downstream audit tasks [48]. Concluding the literature survey and, to the best of our knowledge, this work presents the first step towards learning accounting data representations in a confidentiality-preserving manner.…”

“…Split Learning: When investigating the representations learned by AENs [52] it becomes apparent that representations learned by early (later) layers of the encoder (decoder) function encode detailed attribute characteristics, e.g., a client's journal entry attribute co-occurrence pattern. In contrast, the later encoder (early decoder) function layers learn representations that encode general or 'meta' posting patterns, e.g., a client's prevalent financial accounting principles [47,48]. Derived from this observation and to increase the level of privacy, we apply principles of Split Learning (SL) [20].…”

“…Lately, such models have been trained to accomplish a various downstream audit tasks, e.g. accounting anomaly detection [43,53,68], audit sampling [48,50] or notes analysis [45,54]. Figure 1 illustrates the exemplary application of a deep bottleneck Autoencoder Network (AEN) to detect anomalies in accounting data [49].…”

RESHAPE: Explaining Accounting Anomalies in Financial Statement Audits by enhancing SHapley Additive exPlanations

“…In addition, Nonnenmacher et al [32] and Schreyer et al [40] demonstrated that AENNs could be used to improve audit sampling during an audit process. Furthermore, it was shown that AENNs can be trained in a self-supervised learning setup to detect accounting anomalies and complete additional down-stream audit tasks [38].…”