ABSTRACT
Technology is advancing at a breakneck speed nowadays, and it may be utilized for both good and harm. As a result of the advancement of technology, e-commerce and online transactions have risen in popularity, with the majority of transactions involving credit cards. People may use credit cards to buy now and pay later on both online and offline purchases. It allows customers to purchase without using cash at any store in any country. As the use of credit cards grows, the likelihood of credit card fraud grows as well. The most vulnerable mechanism for fraud is the credit card system. Financial losses are incurred as a result of credit card fraud. Fraudsters strive to come up with new tactics and gimmicks to defraud organizations and customers every year. Read below instructions carefully before writing Document perform these nefarious and criminal acts. For banks and financial institutions, detecting online transaction fraud is the most difficult task. As a result, it is critical for banks and financial institutions to have effective fraud detection systems in place in order to minimize their losses as a result of credit card fraud transactions. Until now, several academics have devised a variety of methods for detecting and reducing these scams. This study proposes a comparison of the Local Outlier Factor and Isolation Forest algorithm using Python and their thorough experimental findings. This will help us predict if the transaction done is fraudulent or not.
• Isolation Forest: The Isolation forest is an unsupervised algorithm used for anomaly detection that randomly selects data. Instead of trying to build a model of normal instances, it explicitly isolates anomalous points in the dataset. It is a very fast algorithm which consumes very less memory. One of the newest ways to discover anomalies is called the Isolation Forests. The algorithm is based on the fact that what anomalies that are few data points and is unique. As a result of these structures, confusing can be found in a method called isolation. This method is very useful and is very different from all available methods. Introduces the use of isolation as a more effective and efficient way to find confusing than the most widely used basic measurement of distance and congestion. In addition, this method is an algorithm with low linear time and low memory requirement. Create a model that performs well with a small number of trees using small specimens of concentrated size, regardless of the size of the data set. Typical machine learning methods often work best when the patterns they are trying to learn are equal, meaning that the same number of positive and negative behaviors exist in the database. How Separate Forests Work the Isolation Forest algorithm distinguishes visuals by randomly selecting a feature and randomly selecting the value of the distinction between the maximum values and the minimum of the selected feature. The logical argument which separate the visual anomaly is easy because only a few conditions are needed to differentiate those conditions from normal observation. On the other hand, separating common perceptions requires additional circumstances. Therefore, confusing points can be calculated as the number of conditions required to differentiate a given observation. The way that the algorithm constructs the separation is by first creating isolation trees, or random decision trees. Then, the score is calculated as the path length to isolate the observation.
• Logistic Regression: Logistic Regression Machine Learning is basically a segmentation algorithm that falls under the supervised category of machine learning algorithms. This means that it traces its roots to the field of Mathematics. A major role of Logistic retrospective study in Machine Learning is to predict the effect of phase-dependent variability from a set of independent variables. In simple terms, categorized variation means a dual binary variable with its data coded in the form of 1 (represents success / yes) or 0 (stands for failure / no). Although Logistic Regression is one of the easiest ways to learn the machine, it has a variety of applications to differentiate problems from spam detection, diabetes prognosis and even cancer detection and credit card fraud.
• Local Outlier Function: Local outlier factor (LOF) is an algorithm used for unsupervised external detection. Produces confusing points representing external data points in the data set. It does this by measuring the deviation of the density of the data point provided in relation to the nearest data points. Area density is determined by measuring distances between neighboring data points (neighbors near k). So for each data point, area density can be calculated. By comparing these we can test which data points are equally robust and have a smaller volume than their neighbors. Those with less weight are considered outsiders.