Solved: What machine learning algorithms are used for fraud detection? Describe and explain them.

Techniques used for fraud detection fall into two primary classes:
a) statistical techniques 
b) artificial intelligence. 
Examples of statistical data analysis techniques are:

1.Data preprocessing techniques for detection, validation, error correction, and filling up of missing or incorrect data.

2.Calculation of various statistical parameters such as averages, quantiles, performance metrics, probability distributions, and so on. For example, the averages may include average length of call, average number of calls per month.

3.Models and probability distributions of various business activities either in terms of various parameters or probability distributions.

4.Computing user profiles.

5.Time-series analysis of time-dependent data.

6.Clustering and classification to find patterns and associations among groups of data.

7.Matching algorithms to detect anomalies in the behavior of transactions or users as compared to previously known models and profiles. 

*** Some forensic accountants specialize in forensic analytics which is the procurement and analysis of electronic data to reconstruct, detect, or otherwise support a claim of financial fraud.
 The main steps in forensic analytics are 
(a) data collection,
(b) data preparation, 
(c) data analysis, and 
(d) reporting

Fraud management is a knowledge-intensive activity. The main AI techniques used for fraud management include:

1.Data mining to classify, cluster, and segment the data and automatically find associations and rules in the data that may signify interesting patterns, including those related to fraud.

2.Expert systems to encode expertise for detecting fraud in the form of rules.

3.Pattern recognition to detect approximate classes, clusters, or patterns of suspicious behavior either automatically (unsupervised) or to match given inputs.

4.Machine learning techniques to automatically identify characteristics of fraud.

5.Neural networks that can learn suspicious patterns from samples and used later to detect them.

A.The COSO Model:

   The COSO “Internal Control – Integrated Framework,” (COSO Model) describes five interrelated components of internal control that provide the foundation for fraud deterrence. These elements of internal control are the means for which the ‘Opportunity’ factors in the Fraud Triangle can be removed to most effectively limit instances of fraud.
*** The five COSO components are:
1. Control Environment:
    The Control environment consists of the actions, policies, and procedures that reflect the overall attitudes of top management, directors and owners of an entity about internal control and its importance to the 
2. Risk Assessment:
    Risk assessment involves the identification of internal and external means that could potentially defeat the organization’s internal control structure, compromise an asset, and conceal the actions from management.It involves identifying as many potential threats as possible, and evaluating them in a way to determine which require action, and the priority for that action.

3. Control Activities:
      Control procedures are also a prime focus area for fraud deterrence engagements; if control procedures are not adequately defined and consistently enforced within the organization, the opportunity for fraud is introduced.
**Control activities generally fall into the five following specific control activities: 
a) adequate separation of duties; 
b) proper authorization of transactions and activities; 
c) adequate documents and records; 
d) physical control over assets and records; and 
e) independent checks on performance 

4. Information & Communication:
*** Information and Communication relates to the flow of information in two directions within the organization. First, information should flow downward to the line functions and provide the best, most accurate information as needed to allow the function to produce the best results possible.
*** Second, information about performance should flow upwards through management, through both formal and informal communication channels, providing objective feedback. Both communication channels must function effectively to safeguard the organization

5. Monitoring:
      Monitoring activities deal with ongoing or periodic assessment of the quality of internal control performance by management to determine that controls are operating as intended and that they are modified as appropriate for changes in conditions

B.LINK ANALYSIS:
         Link analysis has been used for investigation of criminal activity (fraud detection, counterterrorism, and intelligence), computer security analysis, search engine optimization, market research, medical research, and art.

   Link analysis is used for 3 primary purposes
1.Find matches in data for known patterns of interest;
2.Find anomalies where known patterns are violated;
3.Discover new patterns of interest (social network analysis, data mining).

Applications:
1.FBI Violent Criminal Apprehension Program (ViCAP)
2.Iowa State Sex Crimes Analysis System
3.Minnesota State Sex Crimes Analysis System (MIN/SCAP)
4.Washington State Homicide Investigation Tracking System (HITS)