In this course I will cover, how to develop a Credit Card Fraud Detection model to categorize a transaction as Fraud or Legitimate with very high accuracy using different Machine Learning Models. This is a hands on project where I will teach you the step by step process in creating and evaluating a machine learning model. This course will walk you through the initial data exploration and understanding, data analysis, data preparation, model building and evaluation. We will explore RepeatedKFold, StratifiedKFold, Random Oversampler, SMOTE, ADASYN concepts and then use multiple ML algorithms to create our model and finally focus into one which performs the best on the given dataset. I have splitted and segregated the entire course in Tasks below, for ease of understanding of what will be covered.

The impact of machine learning can already be found in companies across virtually all industries. A recent survey by software development firm STX Next discovered that not only have two out of three CTOs reported that machine learning is being used in their organisations but it is also the most popular AI subset. As we enter the New Year, a number of machine learning tools and models are increasing in prominence and usage, with all those who are interested in the field likely to benefit form keeping on top of trends in this area. One of the most intriguing types of machine learning is so-called unsupervised learning. Without the need for human intervention, these algorithms are able to identify unseen patterns and data grouping.

What are the most important unsupervised machine learning algorithms? In this blog post, we will list what we believe to be the top 8. Unsupervised machine learning means that there is no predefined outcome or label for any data point during training. Without a labeled data set, how does one know which algorithm should be used? There are many possible answers to this question and it all depends on the type of problem you need to solve. The goal of this blog post is to help you figure out which unsupervised machine learning algorithm is best for your problem.

Graph kernel approaches have typically been the most popular strategy for graph classification tasks. Graph kernels can be thought of as functions that measure the similarity of two graphs. They allow kernelized learning algorithms like support vector machines to work directly on charts rather than convert them to fixed-length, real-valued feature vectors through feature extraction. In recent years, the use of Graph Neural Networks (GNNs) based on high-performance message-passing neural networks has exploded (MPNNs). As a result, they've grown increasingly popular for graph categorization.

In this paper, we propose a novel nonconvex approach to robust principal component analysis for HSI denoising, which focuses on simultaneously developing more accurate approximations to both rank and column-wise sparsity for the low-rank and sparse components, respectively. In particular, the new method adopts the log-determinant rank approximation and a novel $\ell_{2,\log}$ norm, to restrict the local low-rank or column-wisely sparse properties for the component matrices, respectively. For the $\ell_{2,\log}$-regularized shrinkage problem, we develop an efficient, closed-form solution, which is named $\ell_{2,\log}$-shrinkage operator. The new regularization and the corresponding operator can be generally used in other problems that require column-wise sparsity. Moreover, we impose the spatial-spectral total variation regularization in the log-based nonconvex RPCA model, which enhances the global piece-wise smoothness and spectral consistency from the spatial and spectral views in the recovered HSI. Extensive experiments on both simulated and real HSIs demonstrate the effectiveness of the proposed method in denoising HSIs.

A new approach called ABRF (the attention-based random forest) and its modifications for applying the attention mechanism to the random forest (RF) for regression and classification are proposed. The main idea behind the proposed ABRF models is to assign attention weights with trainable parameters to decision trees in a specific way. The weights depend on the distance between an instance, which falls into a corresponding leaf of a tree, and instances, which fall in the same leaf. This idea stems from representation of the Nadaraya-Watson kernel regression in the form of a RF. Three modifications of the general approach are proposed. The first one is based on applying the Huber's contamination model and on computing the attention weights by solving quadratic or linear optimization problems. The second and the third modifications use the gradient-based algorithms for computing trainable parameters. Numerical experiments with various regression and classification datasets illustrate the proposed method.

Federated learning (FL) provides a high efficient decentralized machine learning framework, where the training data remains distributed at remote clients in a network. Though FL enables a privacy-preserving mobile edge computing framework using IoT devices, recent studies have shown that this approach is susceptible to poisoning attacks from the side of remote clients. To address the poisoning attacks on FL, we provide a \textit{two-phase} defense algorithm called {Lo}cal {Ma}licious Facto{r} (LoMar). In phase I, LoMar scores model updates from each remote client by measuring the relative distribution over their neighbors using a kernel density estimation method. In phase II, an optimal threshold is approximated to distinguish malicious and clean updates from a statistical perspective. Comprehensive experiments on four real-world datasets have been conducted, and the experimental results show that our defense strategy can effectively protect the FL system. {Specifically, the defense performance on Amazon dataset under a label-flipping attack indicates that, compared with FG+Krum, LoMar increases the target label testing accuracy from $96.0\%$ to $98.8\%$, and the overall averaged testing accuracy from $90.1\%$ to $97.0\%$.

Hilsa is the national fish of Bangladesh. Bangladesh is earning a lot of foreign currency by exporting this fish. Unfortunately, in recent days, some unscrupulous businessmen are selling fake Hilsa fishes to gain profit. The Sardines and Sardinella are the most sold in the market as Hilsa. The government agency of Bangladesh, namely Bangladesh Food Safety Authority said that these fake Hilsa fish contain high levels of cadmium and lead which are detrimental for humans. In this research, we have proposed a method that can readily identify original Hilsa fish and fake Hilsa fish. Based on the research available on online literature, we are the first to do research on identifying original Hilsa fish. We have collected more than 16,000 images of original and counterfeit Hilsa fish. To classify these images, we have used several deep learning-based models. Then, the performance has been compared between them. Among those models, DenseNet201 achieved the highest accuracy of 97.02%.

Anomaly detection on attributed networks is widely used in web shopping, financial transactions, communication networks, and so on. However, most work tries to detect anomalies on attributed networks only considering a single interaction action, which cannot consider rich kinds of interaction actions in multi-view attributed networks. In fact, it remains a challenging task to consider all different kinds of interaction actions uniformly and detect anomalous instances in multi-view attributed networks. In this paper, we propose a Graph Convolution based framework, AnomMAN, to detect \textbf{Anom}aly on \textbf{M}ulti-view \textbf{A}ttributed \textbf{N}etworks. To consider the attributes and all interaction actions jointly, we use the attention mechanism to define the importance of all views in networks. Besides, the Graph Convolution operation cannot be simply applied in anomaly detection tasks on account of its low-pass characteristic. Therefore, AnomMAN uses a graph auto-encoder module to overcome the shortcoming and transform it to our strength. According to experiments on real-world datasets, AnomMAN outperforms state-of-the-art models and two variants of our proposed model. Besides, the Accuracy@50 indicator of AnomMAN reaches 1.000 on the dataset, which shows that the top 50 anomalous instances detected by AnomMAN are all anomalous ones.

There are various situations where the need is to group In simple terms, attention mechanism can be thought of an similar texts into same buckets. We do not have enough additional layer somewhere in a network architecture which previous experience or knowledge to run a classification gives the deep learning model extra controlling parameters to algorithm on top of the available data. Clustering is the refine its learning by paying attention to different parts of the fundamental and intuitive solution to such problems.