Android-based mobile devices have attracted a large number of users because they are easy to use and possess wide range capabilities. Because its popularity, Android has become one the most important platforms for attackers launch their nefarious schemes. Due rising sophistication malware obfuscation detection avoidance tactics, many traditional approaches impractical due limited representation Inspired by success deep learning in learning, this article presents an effective improved neural...

MOOCs are online learning environments which many students use, but the success rate of is low. Machine can be used to predict based on how people learn in MOOCs. Predicting performance promote through various methods, such as identifying low-performance or by grouping together. Recent machine has enabled development predictive models, and ensemble method assist reducing variance bias errors associated with single-machine learning. This study uses a two-phase classification model an...

Owing to the exponential proliferation of internet services and sophistication intrusions, traditional intrusion detection algorithms are unable handle complex invasions due their limited representation capabilities unbalanced nature Internet Things (IoT)-related data in terms both telemetry network traffic. Drawing inspiration from deep learning achievements feature extraction learning, this study, we propose an accurate hybrid ensemble framework (HEDLF) protect against obfuscated...

A feedforward neural network ensemble trained through metaheuristic algorithms has been proposed by researchers to produce a group of optimal networks. This method, however, proven be very time-consuming during the optimization process. To overcome this limitation, we propose metaheuristic-based learning algorithm for building an system, resulting in shorter training time. In our master-slave based is employed process heterogeneous networks, which global search operations are executed on...

Multi-label learning with emerging new labels is a practical problem that occurs in data streams and has become an important research issue the area of machine learning. However, existing models for dealing this require high computational times, there still exists lack research. Based on these issues, paper presents incremental kernel extreme multi-label labels, consisting two parts: novelty detector; classifier. The detector free-user-setting threshold parameters was developed to identify...

Unsupervised Extreme Learning Machine (US-ELM) is the one type of neural network which modified from (ELM) for handle clustering problem. Nevertheless, US-ELM has problem with nonfulfillment solution due to K-Mean algorithm was used cluster made accuracy unstable when training many times. In this paper, K-Harmonic mean proposed instead improve and more stable gained called, Harmonic (Harm-ELM) likewise experiment result compared state-of-the-art show that Harm-ELM can overcome 75% all...

Extreme Learning Machine (ELM) model which learn very faster than other neural networks but the solution was not suitable as expected since randomness of input weights and biases may cause to nonfulfillment solution. Flower Pollination (FP-ELM) that it merged by ELM Algorithm (FPA) adjust weight for improve performance output when were calculated. Nonetheless, FP-ELM overfitting more number hidden nodes used. In this paper, Meta (Meta-FPELM) proposed compart weight, calculate combine last...

The most difficult problem with the extreme learning machine is selection of hidden nodes size. proper number predefined through a trial and error approach. convex incremental (CI-ELM) has been proposed to tackle this problem. CI-ELM an constructive neural network universal approximation abilities. However, we have found that some added into layer, may play minor role in network, which results increase complexity. In order avoid shortcoming, propose here improved optimal ridgelet (ICOR-ELM)....

Accurate and reliable wind power forecasting plays a vital role in the operation management of systems. Hence, it has become necessary to research develop high-accuracy model. However, owing highly nonlinear non-stationary patterns time-series, creating model capable predicting such series accurately is both complicated challenging. Aiming at this challenge, paper introduces new decomposition-based hybrid based on multiple decomposition techniques, neural network with random weights (NNRW),...

The Internet of Things (IoT) has gained popularity in recent years by connecting physical objects to the Internet, enabling innovative applications. To facilitate communication low-power and lossy networks (LLNs), IPv6-based routing protocol for LLNs (RPL) is widely used. However, RPL’s lack specified security models makes it vulnerable threats, particularly sinkhole attacks. Existing attack detection techniques suffer from high delays false positives. overcome these limitations, our...

The convolutional autoencoder (CAE) was proposed on neural network (CNN) and denoising (DAE). CAE can address the corrupted input samples high dimensional problem. However, has a shortcoming involving large training timescale because parameters of are commonly tuned by gradient descent (GD) learning method. In order to alleviate this problem, paper fast based extreme machine (ELM), called (FCDA). FCDA, random hidden nodes used reduce dimension data. After that, ELM is reconstruct cleaned...

In recent years, tomato cultivation has dramatically increased, and the competition in market is high. The price determined through maturity classification. tomatoes' generally assessed a visual inspection of color, texture, size, shape, defects. However, quality human sorting may be poor, cost external control labor Transfer learning (TL) an efficient tool involving reuse pre-trained deep network on new problem image classification applications. this study, we investigated performance...