Study on gas-oil ratio prediction considering the influence of imbalance data

Petroleum Science and Technology, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 20

Published: Jan. 1, 2025

When collecting logging data, the lack of gas-oil ratio (GOR) data samples and few measurements lead to reduced accuracy poor performance traditional machine learning algorithms in predicting GORs. This study proposes a new method that combines augmentation meta-learning improve prediction under conditions. Firstly, is collected using multi-component gas technology, dataset enhanced Conditional Table Generative Adversarial Network (CTGAN). Subsequently, 13 derivative parameters were calculated, Principal Component Analysis (PCA) was employed extract features. On this basis, novel approach proposed by combining with Recurrent Neural (RNN) build model (MAML-RNN). The MAML-RNN achieved mean absolute percentage error (MAPE) 0.91 on real 0.69 synthetic dataset. Compared RNN residual neural network (ResNet) models, MAPE 0.93 2.03 1.29 0.29 dataset, respectively.

Language: Английский

---

Enhanced Deep Autoencoder-Based Reinforcement Learning Model with Improved Flamingo Search Policy Selection for Attack Classification

Journal of Cybersecurity and Privacy, Journal Year: 2025, Volume and Issue: 5(1), P. 3 - 3

Published: Jan. 14, 2025

Intrusion detection has been a vast-surveyed topic for many decades as network attacks are tremendously growing. This heightened the need security in networks web-based communication systems advanced nowadays. The proposed work introduces an intelligent semi-supervised intrusion system based on different algorithms to classify accurately. Initially, pre-processing is accomplished using null value dropping and standard scaler normalization. After pre-processing, enhanced Deep Reinforcement Learning (EDRL) model employed extract high-level representations learn complex patterns from data by means of interaction with environment. enhancement deep reinforcement learning made associating autoencoder (AE) improved flamingo search algorithm (IFSA) approximate Q-function optimal policy selection. feature representations, support vector machine (SVM) classifier, which discriminates input into normal attack instances, classification. presented simulated Python platform evaluated UNSW-NB15, CICIDS2017, NSL-KDD datasets. overall classification accuracy 99.6%, 99.93%, 99.42% datasets, higher than existing frameworks.

Language: Английский

---

An optimized intrusion detection system for resource-constrained IoMT environments: enhancing security through efficient feature selection and classification

The Journal of Supercomputing, Journal Year: 2025, Volume and Issue: 81(6)

Published: April 27, 2025

Language: Английский

---

A cloud‐based hybrid intrusion detection framework using XGBoost and ADASYN‐Augmented random forest for IoMT

IET Communications, Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 17, 2024

Abstract Internet of Medical Things have vastly increased the potential for remote patient monitoring, data‐driven care, and networked healthcare delivery. However, connectedness lays sensitive data fragile medical devices open to security threats that need robust intrusion detection solutions within cloud‐edge services. Current approaches modification be able handle practical challenges result from problems with quality. This paper presents a hybrid framework enhances IoMT networks. There are three modules in design. First, an XGBoost‐based noise model is used identify anomalies. Second, adaptive resampling ADASYN done fine‐tune class distribution address imbalance. Third, ensemble learning performs through Random Forest classifier. stacked coordinates techniques filter preprocess imbalanced data, identifying high accuracy reliability. These results then experimentally validated on UNSW‐NB15 benchmark demonstrate effective under realistically noisy conditions. The novel contributions work new structural paradigm coupled integrated filtering, learning. proposed advanced oversampling gives performance surpasses all others reported 92.23% accuracy.

Language: Английский

---

Optimizing UPVC profile production using adaptive neuro-fuzzy inference system

International Journal of Information Technology, Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 27, 2024

Language: Английский

---

A Novel Few-Shot ML Approach for Intrusion Detection in IoT

Arabian Journal for Science and Engineering, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 6, 2024

Language: Английский

---

Ensemble feature selection and tabular data augmentation with generative adversarial networks to enhance cutaneous melanoma identification and interpretability

BioData Mining, Journal Year: 2024, Volume and Issue: 17(1)

Published: Oct. 30, 2024

Cutaneous melanoma is the most aggressive form of skin cancer, responsible for cancer-related deaths. Recent advances in artificial intelligence, jointly with availability public dermoscopy image datasets, have allowed to assist dermatologists identification. While feature extraction holds potential detection, it often leads high-dimensional data. Furthermore, datasets present class imbalance problem, where a few classes numerous samples, whereas others are under-represented.

Language: Английский

---

Forecasting the Metal Ores Industry Index on the Tehran Stock Exchange: A Gated Recurrent Unit (GRU) Approach

Journal of Artificial Intelligence and Capsule Networks, Journal Year: 2024, Volume and Issue: 6(4), P. 436 - 451

Published: Nov. 16, 2024

This research offers an in-depth examination of predicting the closing prices metal ores industry index on Tehran Stock Exchange (TSE) using a Gated Recurrent Unit (GRU) model. The GRU, type recurrent neural network, shows great promise for tasks involving time series forecasting. historical daily price data from October 2017 to 2022, was used in study after carefully preprocessing it further analysis. begins with univariate analysis reveal distribution characteristics and relationships between essential variables. A customized GRU model that is trained 70% data, its performance assessed through metrics such as Root Mean Square Error (RMSE), (MSE), Absolute (MAE), R-squared (R2) score prediction. results indicate provides accurate predictions index, outperforming traditional forecasting techniques. model's nature enables capture both short-term long-term temporal dependencies within data. highlights significant potential networks realm financial Future improvements will focus hyperparameter optimization integrating additional input variables enhance predictive accuracy.

Language: Английский

---