Digital Transformation of Financial Services using Artificial Intelligence, Machine Learning, and Cloud Computing

Digital disruption is redefining industries and changing the way business function. Artificial Intelligence is the future of banking as it brings the power of advanced data analytics to combat fraudulent transactions and improve compliance. Financial services are the economical backbone of any nation in the world. There are billions of financial transactions which are taking place and all this data is stored and can be considered as a gold mine of data for many different organizations. No human intelligence can dig in this amount of data to come up with something valuable. This is the reason financial organizations are employing artificial intelligence to come up with new algorithms which can change the way financial transactions are being carried out. Artificial Intelligence can complete the task in a very short period. Artificial intelligence can be used to detect frauds, identify possible attacks, and any other kind of anomalies that may be detrimental for the institution. This paper discusses the role of artificial intelligence and machine learning in the finance sector. Additionally, the paper will provide the necessary strategies that any banking organization can follow when digitizing its operations when implementing Artificial Intelligence, Machine learning and Cloud Computing.

Category Trees - Classifiers that Branch on Category

This paper presents a batch classifier that splits a dataset into tree branches depending on the category type. It has been improved from the earlier version and fixed a mistake in the earlier paper. Two important changes have been made. The first is to represent each category with a separate classifier. Each classifier then classifies its own subset of data rows, using batch input values to create the centroid and also represent the category itself. If the classifier contains data from more than one category however, it needs to create new classifiers for the incorrect data. The second change therefore is to allow the classifier to branch to new layers when there is a split in the data, and create new classifiers there for the data rows that are incorrectly classified. Each layer can therefore branch like a tree - not for distinguishing features, but for distinguishing categories. The paper then suggests a further innovation, which is to represent some data columns with fixed value ranges, or bands. When considering features, it is shown that some of the data can be classified directly through fixed value ranges, while the rest must be classified using a classifier technique and the idea allows the paper to discuss a biological analogy with neurons and neuron links. Tests show that the method can successfully classify a diverse set of benchmark datasets to better than the state-of-the-art.

Automation of Best-Fit Model Selection using a Bag of Machine Learning Libraries for Sales Forecasting

Sales forecasting became crucial for industries in past decades with rapid globalization, widespread adoption of information technology towards e-business, understanding market fluctuations, meeting business plans, and avoiding loss of sales. This research precisely predicts the automotive industry sales using a bag of multiple machine learning and time series algorithms coupled with historical sales and auxiliary features. Three-year historical sales data (from 2017 till 2020) were used for the model building or training, and one-year (2020-2021) predictions were computed for 900 unique SKU's (stock-keeping units). In the present study, the SKU is a combination of sales office, core business field, and material customer group. Various data cleaning and exploratory data analysis algorithms were implemented over raw datasets before use for modeling. Mean absolute percentage error (mape) were estimated for individual predictions from time series and machine learning models. The best model was selected for unique SKU's as per the most negligible mape value.

Automatic Home-based Screening of Obstructive Sleep Apnea using Single Channel Electrocardiogram and SPO2 Signals

Obstructive sleep apnea (OSA) is one of the most widespread respiratory diseases today. Complete or relative breathing cessations due to upper airway subsidence during sleep is OSA. It has confirmed potential influence on Covid-19 hospitalization and mortality, and is strongly associated with major comorbidities of severe Covid-19 infection. Un-diagnosed OSA may also lead to a variety of severe physical and mental side-effects. To score OSA severity, nocturnal sleep monitoring is performed under defined protocols and standards called polysomnography (PSG). This method is time-consuming, expensive, and requiring professional sleep technicians. Automatic home-based detection of OSA is welcome and in great demand. It is a fast and effective way for referring OSA suspects to sleep clinics for further monitoring. On-line OSA detection also can be a part of a closed-loop automatic control of the OSA therapeutic/assistive devices. In this paper, several solutions for online OSA detection are introduced and tested on 155 subjects of three different databases. The best combinational solution uses mutual information (MI) analysis for selecting out of ECG and SpO2-based features. Several methods of supervised and unsupervised machine learning are employed to detect apnoeic episodes. To achieve the best performance, the most successful classifiers in four different ternary combination methods are used. The proposed configurations exploit limited use of biological signals, have online working scheme, and exhibit uniform and acceptable performance (over 85%) in all the employed databases. The benefits have not been gathered all together in the previous published methods.

A Knowledge based Automatic Radiation Treatment Plan Alert System

In radiation therapy, preventing treatment plan errors is of paramount importance. In this paper, an alert system is proposed and developed for checking if the pending cancer treatment plan is consistent with the intended use. A key step in the development of the paper is characterization of various treatment plan fingerprints by three-dimension vectors taken from possibly thousands of variables in each treatment plan. Then three machine learning based algorithms are developed and tested in the paper. The first algorithm is a knowledge-based support vector machine method. If an incorrect treatment plan were offered, the algorithm would tell that the pending treatment plan is inconsistent with the intended use and provide a red flag. The algorithm is tested on the actual patient data sets with 100% successful rate and 0% failure rate. In addition, two algorithms based on the well-known k-nearest neighbour and Bayesian approach respectively are developed. Similar to the support vector machine algorithm, these two algorithms are also tested with 100% success rate and 0% failure rate. The key seems to pick up the right features.

A New Perspective of Paramodulation Complexity by Solving 100 Sliding Block Puzzles

This paper gives complete guidelines for authors submitting papers for the AIRCC Journals. A sliding puzzle is a combination puzzle where a player slides pieces along specific routes on a board to reach a certain end configuration. In this paper, we propose a novel measurement of the complexity of 100 sliding puzzles with paramodulation, which is an inference method of automated reasoning. It turned out that by counting the number of clauses yielded with paramodulation, we can evaluate the difficulty of each puzzle. In the experiment, we have generated 100 * 8 puzzles that passed the solvability checking by countering inversions. By doing this, we can distinguish the complexity of 8 puzzles with the number generated with paramodulation. For example, board [2,3,6,1,7,8,5,4, hole] is the easiest with score 3008 and board [6,5,8,7,4,3,2,1, hole] is the most difficult with score 48653.Besides, we have succeeded in obverse several layers of complexity (the number of clauses generated) in 100 puzzles. We can conclude that the proposed method can provide a new perspective of paramodulation complexity concerning sliding block puzzles.

An Acceleration Method based on Deep Learning and Multilinear Feature Space

Computer vision plays a crucial role in Advanced Assistance Systems. Most computer vision systems are based on Deep Convolutional Neural Networks (deep CNN) architectures. However, the high computational resource to run a CNN algorithm is demanding. Therefore, the methods to speed up computation have become a relevant research issue. Even though several works on architecture reduction found in the literaturehave not yet been achievedsatisfactory results for embedded real-time system applications. This paper presents an alternative approach based on the Multilinear Feature Space (MFS) method resorting to transfer learning from large CNN architectures. The proposed method uses CNNs to generate feature maps, although it does not work as complexity reduction approach. After the training process, the generated features maps are used to create vector feature space. We use this new vector space to make projections of any new sample to classify them. Our method, named AMFC, uses the transfer learning from pre-trained CNN to reduce the classification time of new sample image, with minimal accuracy loss. Our method uses the VGG-16 model as the base CNN architecture for experiments; however, the method works with any similar CNN model. Using the well-known Vehicle Image Database and the German Traffic Sign Recognition Benchmark, we compared the classification time of the original VGG-16 model with the AMFCmethod, and our method is, on average, 17 times faster. The fast classification time reduces the computational and memory demands in embedded applications requiring a large CNN architecture.

Finding Facial Expression Patterns on Videos based on Smile and Eyes-Open Confidence Values

Facial expression recognition is one of the types of non-verbal communication that is not only commons for human but also plays an essential role in everyday lives. The development of science and technology allows the machine to automatically detect human facial expressions based on images and videos. Numerous facial expression detection methods have been proposed in the literature. This paper presents a method to find three basic facial expressions (neutral, happy, and angry) from two parameter values: smile and eyes-open. The analysis involves a preprocessing step using a combination of pre-designed proprietary algorithm and Luxand library. Firstly, the parameters were mapped into two-dimensional space and then grouped into three clusters using K-means, a popular heuristic clustering method. Secondly, more than 50,000 frames for each video were experimented using the proprietary research data. The result shows that the proposed method successfully performed a simple video analysis of facial expressions.

A Brief Survey of Question Answering Systems

Question Answering (QA) is a subfield of Natural Language Processing (NLP) and computer science focused on building systems that automatically answer questions from humans in natural language. This survey summarizes the history and current state of the field and is intended as an introductory overview of QA systems. After discussing QA history, this paper summarizes the different approaches to the architecture of QA systems -- whether they are closed or open-domain and whether they are text-based, knowledge-based, or hybrid systems. Lastly, some common datasets in this field are introduced and different evaluation metrics are discussed.

Human-System Interface with Explanation of Actions for Autonomous Anti-UAV Systems

Research on explanation is currently of intense interest as documented in the DARPA 2021 investments reported by the USA Department of Defense. An emerging theme for explanation techniques research is their application to the improvement of human-system interfaces for autonomous anti-drone or C-UAV defense systems. In the present paper a novel proposal based on natural language processing technology concerning explanatory discourse using relations is briefly described. The proposal is based on the use of relations pertaining to the possible malicious actions of an intruding alien drone swarm and the defense decisions proposed by an autonomous anti-drone system. The aim of such an interface is to facilitate the supervision that a user must exercise on an autonomous defense system in order to minimize the risk of wrong mitigation actions and unnecessary spending of ammunition.