INTEGRATING MACHINE LEARNING ALGORITHMS WITH QUANTUM ANNEALING SOLVERS FOR ONLINE FRAUD DETECTION
ABSTARCT :
Machine learning has been increasingly applied in identification of fraudulent transactions. However, most application systems detect duplicitous activities after they have already occurred, not at or near real time. Since spurious transactions are far fewer than the normal ones, the highly imbalanced data makes fraud detection very challenging and calls for ways to address it beyond the traditional machine learning approach. This study has proposed a detection framework, and implemented it using quantum machine learning (QML) approach by applying Support Vector Machine (SVM) enhanced with quantum annealing solvers
EXISTING SYSTEM :
? Secondly, in spite of the need for real time or near real time fraud detection for online e-commerce transactions, the effectiveness of many existing systems is compromised since most detect only after the fraud activities have happenedwhen the loss has already occurred.
? Besides the models designed for directly analyzing non-stationary data with the ‘‘time’’ variable, there are existing approaches to transform non-stationary to stationary data by detrending methods such as power transform, square root, and log transform
? Because of quantum computing’s powerful modelling abilities to solve some complex problems that existing computing cannot, we consider QML a promising approach to tackle the huge volume of online fraud data
DISADVANTAGE :
? Because of quantum computing’s powerful modelling abilities to solve some complex problems that existing computing cannot, we consider QML a promising approach to tackle the huge volume of online fraud data. Details on why QML are used in this study are discussed
? It is a supervised machine learning method for two-group classification problems.
? More complex kernel functions can be obtained by solving quadratic constrained binary optimization problem [27], and that require very high computing capability. One solution is to develop a general quadratic constrained model for SVM, and recast it explicitly as a quadratic unconstrained binary optimization problem (QUBO) using quadratic infeasibility penalties as constraints .
PROPOSED SYSTEM :
? This study has proposed a detection framework, and implemented it using quantum machine learning (QML) approach by applying Support Vector Machine (SVM) enhanced with quantum annealing solvers.
? shows the proposed fraud detection framework and its components.
? Although machine learning algorithms have been proposed, they are still under the assumption of stationary or non-time series data.
ADVANTAGE :
? Describes the characteristics of the datasets used and the algorithms evaluated in this study.
? Once the optimal hyperplane is constructed, it is then used to separate the transactions into normal and fraudulent groups.
? Augmented Dickey Fuller (ADF) and Kwiatkowski-Phillips-Schmidt-Shin (KPSS) are two of the most commonly used statistical test to analyze whether the series of data are the stationary, this study uses both tests to evaluate whether the time series data is stationary as shown in via the unit root test
|