Impact of Data Balancing During Training for Best Predictions

In this paper, an investigation was made to evaluate the effectiveness of the different classifiers suitable to predict the probability of a cyber-threat or fraudulent intent applicant during the Mobile Money Service on-boarding or service activation process, with the goal of determining the best machine learning model for the predictive model solution. Experimental work was carried out by formulating cyber threat predictive models using six supervised machine learning algorithms: Logistic regression(LR), Naïve Bayes, Shallow Neural Network (SNN), Deep Neural Network (DNN), Classification and Regression Trees (CART) and Random Forest (RF) of different configurations. Each model was simulated with both Synthetic Minority Operation Techniques (SMOTE) and without SMOTE (No-SMOTE) on 25,000 records of mobile money applicants. Twenty-four ( 24) different configurations of the formulated predictive models were simulated and evaluated using the Python programming language. Simulation results of the predictive models proved that the Random Forest model multiclass configurations with the SMOTE dataset outperformed all other configurations. The results also showed that the multiclass experiments with SMOTE had better performance than the binary configurations with NO-SMOTE in the predictive models. The study concluded that using the Random Forest-based predictive machine learning model will increase the security level of the Mobile Money solution by detecting and preventing anomalous customer registrations during the unbanked onboarding process.Povzetek: Napovedovanje kibernetskih groženj mobilnega denarja z uporabo algoritmov strojnega učenja. 22 RF Binary-SMOTE Random Forest(RF) with Binary feature configuration and with SMOTE application to Dataset 23 RF Multiclass-No SMOTE Random Forest(RF) with Multiclass feature configuration and No-SMOTE application to Dataset 24 RF Multiclass-SMOTE Random Forest(RF) with Multiclass feature configuration and with SMOTE application to Dataset

Section: Related Workmentioning

confidence: 99%

Performance Evaluation of Machine Learning Models for Cyber Threat Detection and Prevention in Mobile Money Services

Akinyemi

Olalere²,

Sanni

et al. 2023

“…Synthetic Minority Over-sampling Technique (SMOTE) algorithm is one of the well-known augmentation techniques that are used in imbalanced datasets to solve minority class problems. In the imbalanced dataset, there are too few instances of minority classes that affect model decisions [25].…”

Section: Using Data Augmentation Techniquementioning

confidence: 99%

Performance of Malware Detection Classifier Using Genetic Programming in Feature Selection

Harahsheh¹,

Shraideh²,

Sharaeh³

2021

The term "malicious software," which is commonly referred to as malware, describes malicious software that affects or harms computers, servers, or networks. While the numbers and complexity of malware have rapidly increased, developing a malware detection system is required to detect malware in the world of cybersecurity and test the behavior of its new features. While traditional techniques provide less efficiency in detecting new malware, machine learning techniques are used to achieve rapid malware detection in an intelligent way to improve detection performance, as malware and its application in the industry are constantly increasing. In this study, we developed a malware detection model by detecting malware using machine learning classifiers, after passing a new feature selection technique using genetic programming. We also compared the performance of all classifiers using the most recent feature selection techniques. Results show that Random Forest, Random Forest ( 4), and Random Tree give the best value in all experiments, while Hoeffding Tree and Decision Stump give lower values for F1-score and accuracy in all experiments. The feature selection method that proposed GPMP gives a better value than Filter-based with little differences. The accuracy and F1-score have the values of 0.881066 and 0.867546 for GPMP, and the values of 0.877624 and 0.862894 for Filter-based, respectively. The experimental results reveal that GPMP used fewer features than Filter-based, and this affected the computation and complexity of the model. Povzetek: Analizirane so bile metode strojnega učenja za povečanje uspešnosti odkrivanja zlonamerne programske opreme.

“…Given any dataset on financial transactions such as MMTs, the number of fraudulent transactions (positive class) compared to the legitimate (negative class) ones, constitutes a very small percentage of the dataset. This makes the datasets highly imbalanced [6] and predictions from such data using machine learning algorithms are skewed towards the legitimate transactions with the long term effect that predictions made with such data can be misleading. We select Logistic Regression as the machine learning algorithm for this work as it has proven its potency [7] in a multitude of fields for classification and prediction.…”

Section: Introductionmentioning

confidence: 99%

Predicting Fraud in Mobile Money Transactions using Machine Learning: The Effects of Sampling Techniques on the Imbalanced Dataset

Botchey

Qin

Hughes-Lartey

et al. 2022

Mobile Money Fraud is advancing in developing countries. We propose a solution to this problem based on machine learning. Labeled data from financial transactions which includes mobile money transactions are however, skewed towards the legitimate transactions. Machine learning models built with such skewed datasets are unreliable as the prediction algorithms will be biased towards the legitimate transactions. We investigate the performance of different sampling and weighting techniques such as Adaptive Synthetic Sampling (ADASYN) and Synthetic Minority Oversampling Technique (SMOTE). We select Logistic Regression for the experiments due to its simplicity and relatively low computational needs. The performance is evaluated with different metrics. Manually tuning the weights of the classes achieved the best results in our experiments.Povzetek: Opisana je metoda za detekcijo prevar v mobilnih transakcijah s pomočjo strojnega učenja na neuravnoteženih podatkih.