UAS Mata Kuliah Metodologi Penelitian - Presentasi Seminar Proposal Skripsi - Meakhel Gunawan

Summary

Michael Gunwan presents a seminar proposal on detecting credit card fraud using a combination of Smooth SVM and feature selection techniques. The research addresses the challenge of imbalanced data, where normal transactions vastly outnumber fraudulent ones. By applying Smooth SVM to balance the data and feature selection to identify relevant variables, the proposal aims to improve fraud detection accuracy. The methodology involves data collection, preprocessing, oversampling, and model evaluation using SVM, with performance metrics such as accuracy, precision, and recall to assess the effectiveness of the model in real-world applications.

Takeaways

• 😀 The speaker, Michael Gunwan, is presenting a research proposal on detecting credit card fraud using a combination of SMOTE, feature selection, and SVM.
• 😀 The research aims to address the growing issue of financial fraud, particularly in the context of credit card transactions, which are increasingly automated and easy to access.
• 😀 A key challenge in fraud detection is the imbalance in data, with normal transactions far outnumbering fraudulent ones, making it difficult for machine learning models to recognize fraud patterns.
• 😀 The hypothesis suggests that combining SMOTE, feature selection, and SVM can improve the accuracy of fraud detection by addressing data imbalance and optimizing model classification.
• 😀 Previous research in the field has shown both successes and limitations of different approaches, such as neural networks and SVM, in detecting fraud on imbalanced datasets.
• 😀 SMOTE (Synthetic Minority Over-sampling Technique) will be used to balance the dataset by oversampling the minority fraud class to make it comparable with the majority normal transactions.
• 😀 Feature selection will be conducted using filter methods (correlation analysis) and embedded methods (feature importance) to identify the most relevant variables for fraud detection.
• 😀 The SVM model will be trained and evaluated using various metrics, including accuracy, precision, recall, and F1 score, to assess its effectiveness in detecting fraud.
• 😀 The dataset for this study consists of 284,807 transactions, where only 0.17% are fraudulent, highlighting the severe class imbalance.
• 😀 The research methodology includes data preprocessing, feature selection, SMOTE application, SVM modeling, and iterative refinement of hyperparameters to optimize the fraud detection model.

Q & A

What is the main topic of Michael Gunwan's seminar proposal?

The main topic is the combination of SMO SVM and feature selection for fraud detection in imbalanced credit card transaction data.

Why is fraud detection in the financial sector important, according to the speaker?

Fraud detection is critical due to the rise in financial fraud, particularly in the form of cybercrime, which leads to significant financial losses. In 2024, financial fraud losses reached 2.5 trillion, highlighting the need for effective detection mechanisms.

What challenge does the imbalance of data present in fraud detection models?

Data imbalance, where normal transactions vastly outnumber fraudulent ones, makes it difficult for models to recognize rare fraud patterns. This imbalance reduces the accuracy of fraud detection models as they tend to favor the majority class (normal transactions).

What is the hypothesis presented by the speaker regarding fraud detection?

The hypothesis suggests that data imbalance negatively affects fraud detection accuracy. By using a combination of SMO, feature selection, and SVM, it is expected that the model's performance will improve by balancing the data, selecting relevant features, and optimizing classification.

What methods were used in the literature to detect fraud, and what were their limitations?

Previous research used methods like neural networks and SVM. However, these methods were often limited by issues such as overfitting and high computational resource requirements, making them less suitable for large-scale fraud detection tasks.

What is SMO, and how does it help in handling imbalanced data?

SMO (Sequential Minimal Optimization) is an oversampling technique that adds more instances to the minority class (fraudulent transactions) to balance the dataset. This helps in preventing the model from being biased towards the majority class (normal transactions), improving its ability to detect fraud.

How does feature selection contribute to the fraud detection model?

Feature selection helps by identifying and retaining the most relevant variables for fraud detection. It reduces dimensionality by eliminating irrelevant or redundant features, which can improve the model's efficiency and performance.

What type of dataset is used in the fraud detection model, and what is its key characteristic?

The dataset used contains 284,807 transaction records, with two classes: fraud (0.17%) and non-fraud (99.83%). The key characteristic is the extreme imbalance between the two classes, where fraudulent transactions are much fewer compared to normal transactions.

What evaluation metrics are used to assess the performance of the fraud detection model?

The model is evaluated using accuracy, precision, recall, F1 score, and confusion matrix. These metrics provide insights into the model's effectiveness in identifying fraudulent transactions and its ability to minimize false positives and false negatives.

What steps are involved in the methodology of the research?

The methodology includes data collection and preprocessing, feature selection, applying SMO to balance the data, training an SVM model, evaluating the model's performance using various metrics, and iterating with hyperparameter adjustments if needed to improve results.