Fighting Money Laundering With Statistics and

Machine Learning
ABSTRACT

Money laundering is a profound global problem. Nonetheless, there is little scientific

literature on statistical and machine learning methods for anti-money laundering. In
this paper, we focus on anti-money laundering in banks and provide an introduction
and review of the literature. We propose a unifying terminology with two central
elements: (i) client risk profiling and (ii) suspicious behavior flagging.We find that
client risk profiling is characterized by diagnostics, i.e., efforts to find and explain
risk factors. On the other hand, suspicious behavior flagging is characterized by non-
disclosed features and hand-crafted risk indices. Finally, we discuss directions for
future research. One major challenge is the need for more public data sets. This may
potentially be addressed by synthetic data generation. Other possible research
directions include semi-supervised and deep learning, interpretability, and fairness of
the results.

EXISTING SYSTEM

Badal-Valero et al. [37] combine Benford’s Law and four machine learning models.
Benford’s Law [38] gives an empirical distribution of leading digits. The authors use
it to extract features from financial statements. Specifically, they consider statements
from 335 suppliers to a company on trial for money laundering. Of these, 23
suppliers have been investigated and labeled as colluders. All other (non-
investigated) suppliers are treated as benevolent. The motivating idea is that any
colluders, hiding in the non-investigated group, should be misclassified by the
employed models. These include a logistic regression, feedforward neural network,
decision tree, and random forest. Random forests [39], in particular, combine
multiple decision trees. Every tree uses a random subset of features in every node
split. To address class imbalance, i.e., the unequal distribution of labels, the authors
investigate weighting and synthetic minority oversampling [40]. The former weighs
observations during training, giving higher importance to data from the minority
class. The latter balances the data before training, generating synthetic observations
of the minority class. According to the authors, synthetic minority oversampling
works the best. However, the conclusion is apparently based on simulated evaluation
data.

González and Valásquez [41] employ a decision tree, feedforward neural network,
and Bayesian network to model Chilean firms using false invoices. Bayesian
networks [42], in particular, are probabilistic models that represent variable
dependencies via directed acyclic graphs. The authors use data on 582,161 firms,
1,692 of which have been labeled as either fraudulent or non-fraudulent. Features
include information about previous audits and taxes paid. Because most firms are
unlabeled, the authors first use unsupervised learning to characterize high-risk
behavior. To this end, they employ self-organizing maps [43] and neural gas [44].
Both are neural network techniques that build on competitive learning [45] rather
than error correction (i.e., gradient-based optimization). While the methods do
produce clusters with some behavioral patterns, they do not appear useful for false
invoice detection. On the labeled training data, the feedforward neural network
achieves the best performance.

Camino et al. [58] flag clients with three outlier detection techniques: an isolation
forest, a one-class support vector machine, and a Gaussian mixture model. Isolation
forests [59] build multiple decision trees using random feature splits. Observations
isolated by comparatively few feature splits (averaged over all trees) are then
considered outliers. One-class support vector machines [60] use a kernel function to
map data into a reproducing Hilbert space. The method then seeks a maximum
margin hyperplane that separates data points from the origin. A small number of
observations are allowed to violate the hyperplane; these are considered outliers.
Finally, Gaussian mixture models [61] assume that all observations are generated by
a number of Gaussian distributions. Observations in low-density regions are then
considered outliers. The authors combine all three techniques into a single ensemble
method. The method is tested on a data set from an AML software company. This
contains one million transactions with client-level features recording summary
statistics. The authors report positive feedback from the data-supplying company;
otherwise, evaluation is limited.

Sun et al. [62] apply extreme value theory [63] to flag outliers in transaction streams.
The authors start by engineering two features. The first records the number of times
an account has reached a balanced state, i.e., when money transferred into an account
is transferred out again. The second records the number of effective fan-ins
associated with an account, i.e., when money transferred into the account surpasses a
given limit and the account again reaches a balanced state. Next, the Pickands–
Balkema–De Haan theorem [64], [65] is invoked to model (derived) conditional
feature exceedances according to a generalized Pareto distribution. The approach
allows the authors to flag transactions according to a probabilistic limit p (in analogy
to the p-values used to test null hypotheses). The approach is tested on real bank data
with simulated noise and outliers.

Disadvantages
 We find that studies on client risk profiling are characterized by diagnostics,
i.e., efforts to find and explain risk factors. Specifically, unsupervised methods
are used to search for new ‘‘risky’’ observations or risk factors. On the other
hand, supervised methods are used with an explanatory focus.
  We also find that studies employing unsupervised methods generally use
relatively large data sets. By contrast, studies employing supervised methods
use small (labeled) data sets.

Proposed System

In this paper, we focus onAMLin banks and aim to provide a technical review that
researchers and industry practitioners (statisticians and machine learning engineers)
can use as a guide to the current literature on statistical and machine learning
methods for AML in banks. Furthermore, we aim to provide a terminology that can
facilitate policy discussions, and to provide guidance on open challenges within the
literature. To achieve our aims, we (i) propose a unified terminology for AML in
banks, (ii) review selected exemplary methods, and (iii) present recent machine

learning concepts that may improve AML.

Advantages

 The proposed system reduced an UNSUPERVISED CLIENT RISK

PROFILING problem.
 The proposed system eliminates SUPERVISED CLIENT RISK PROFILING
problem.

SYSTEM REQUIREMENTS

➢ H/W System Configuration:-

➢ Processor - Pentium –IV

➢ RAM - 4 GB (min)
➢ Hard Disk - 20 GB
➢ Key Board - Standard Windows Keyboard
 ➢ Mouse - Two or Three Button Mouse
➢ Monitor - SVGA

SOFTWARE REQUIREMENTS:

 Operating system : Windows 7 Ultimate.

 Coding Language : Python.

 Front-End : Python.

 Back-End : Django-ORM

 Designing : Html, css, javascript.

 Data Base : MySQL (WAMP Server).