Have a personal or library account? Click to login
Comparative study of deep learning explainability and causal ai for fraud detection Cover

Comparative study of deep learning explainability and causal ai for fraud detection

Open Access
|Aug 2024

Figures & Tables

Figure 1:

Research methodology flowchart.
Research methodology flowchart.

Figure 2:

Confusion matrix of the (A) first and (B) second deep learning models.
Confusion matrix of the (A) first and (B) second deep learning models.

Figure 3:

LIME explanation for single instance for (A) first and (B) second deep learning model. LIME, local interpretable model-agnostic explanations.
LIME explanation for single instance for (A) first and (B) second deep learning model. LIME, local interpretable model-agnostic explanations.

Figure 4:

Force plot of the (A) first deep learning model and (B) second deep learning model.
Force plot of the (A) first deep learning model and (B) second deep learning model.

Figure 5:

Summary plot of the first deep learning model.
Summary plot of the first deep learning model.

Figure 6:

Summary plot of the second deep learning model.
Summary plot of the second deep learning model.

Figure 7:

(A) First SCM; (B) Second SCM; and (C) Third SCM. SCM, structured causal model.
(A) First SCM; (B) Second SCM; and (C) Third SCM. SCM, structured causal model.

Figure 8:

(A) SCM of the first new data; (B) SCM of the second new data; and (C) SCM of the third new data. SCM, structured causal model.
(A) SCM of the first new data; (B) SCM of the second new data; and (C) SCM of the third new data. SCM, structured causal model.

Layers of the first deep learning model

Model: “sequential”

Layer (type)Output shapeParam #
conv1d (Conv1D)(None, 29, 64)256
max_pooling1d (MaxPooling1D)(None, 14, 64)0
flatten (Flatten)(None, 896)0
dense (Dense)(None, 64)57,408
dense_1 (Dense)(None, 1)65

Total params: 57,729
Trainable params: 57,729
Non-trainable params: 0

Layers of the second deep learning model

Model: “sequential”

Layer (type)Output shapeParam #
conv1d (Conv1D)(None, 27, 128)768
max_pooling1d (MaxPooling1D)(None, 13, 128)0
flatten (Flatten)(None, 1664)0
dense (Dense)(None, 128)213,120
dense_1 (Dense)(None, 1)129

Total params: 214,017
Trainable params: 214,017
Non-trainable params: 0

Evaluation matrix of the BNs

Accuracy (%)Precision (%)Recall (%)F1 score (%)
SCM 149.3255.6650.7936.59
SCM 259.2760.5059.0257.67
SCM 354.9463.0055.6547.95

Different synthetic BAF tabular datasets

BaseSampled to best represent original dataset
Variant IHas higher group size disparity than base
Variant IIHas higher prevalence disparity than base
Variant IIIHas better separability for one of the groups
Variant IVHas higher prevalence disparity in train
Variant VHas better separability in train for one of the groups

Evaluation matrix of the deep learning models

Accuracy (%)Precision (%)Recall (%)F1 score (%)
Model 194.6497.4297.7094.47
Model 296.3496.0696.6496.35

Evaluation matrix of the BN for new data

No. of columnsAccuracy (%)Precision (%)Recall (%)F1 score (%)
New Data 11090.3690.6991.6790.32
New Data 2878.478.478.478.4
New Data 31177.9477.9477.9477.94
Language: English
Submitted on: Apr 25, 2024
Published on: Aug 6, 2024
Published by: Professor Subhas Chandra Mukhopadhyay
In partnership with: Paradigm Publishing Services
Publication frequency: 1 issue per year

© 2024 Erum Parkar, Shilpa Gite, Sashikala Mishra, Biswajeet Pradhan, Abdullah Alamri, published by Professor Subhas Chandra Mukhopadhyay
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.