Search results

Automatic anatomical therapeutic chemical (ATC) classification is progressing at a rapid pace because of its potential in drug development. Predicting an unknown compound's therapeutic and chemical characteristics in terms of how it affects multiple organs and physiological systems makes automatic ATC classification a vital yet challenging multilabel problem. The aim of this paper is to experimentally derive an ensemble of different feature descriptors and classifiers for ATC classification that outperforms the state-of-the-art.

The proposed method is an ensemble generated by the fusion of neural networks (i.e. a tabular model and long short-term memory networks (LSTM)) and multilabel classifiers based on multiple linear regression (hMuLab). All classifiers are trained on three sets of descriptors. Features extracted from the trained LSTMs are also fed into hMuLab. Evaluations of ensembles are compared on a benchmark data set of 3883 ATC-coded pharmaceuticals taken from KEGG, a publicly available drug databank.

Experiments demonstrate the power of the authors’ best ensemble, EnsATC, which is shown to outperform the best methods reported in the literature, including the state-of-the-art developed by the fast.ai research group. The MATLAB source code of the authors’ system is freely available to the public at https://github.com/LorisNanni/Neural-networks-for-anatomical-therapeutic-chemical-ATC-classification.

This study demonstrates the power of extracting LSTM features and combining them with ATC descriptors in ensembles for ATC classification.

In this paper, we present a study about an automated system for monitoring underwater ecosystems. The system here proposed is based on the fusion of different deep learning methods. We study how to create an ensemble based of different Convolutional Neural Network (CNN) models, fine-tuned on several datasets with the aim of exploiting their diversity. The aim of our study is to experiment the possibility of fine-tuning CNNs for underwater imagery analysis, the opportunity of using different datasets for pre-training models, the possibility to design an ensemble using the same architecture with small variations in the training procedure.

Our experiments, performed on 5 well-known datasets (3 plankton and 2 coral datasets) show that the combination of such different CNN models in a heterogeneous ensemble grants a substantial performance improvement with respect to other state-of-the-art approaches in all the tested problems. One of the main contributions of this work is a wide experimental evaluation of famous CNN architectures to report the performance of both the single CNN and the ensemble of CNNs in different problems. Moreover, we show how to create an ensemble which improves the performance of the best single model. The MATLAB source code is freely link provided in title page.

Diagnosing pain in neonates is difficult but critical. Although approximately thirty manual pain instruments have been developed for neonatal pain diagnosis, most are complex, multifactorial, and geared toward research. The goals of this work are twofold: 1) to develop a new video dataset for automatic neonatal pain detection called iCOPEvid (infant Classification Of Pain Expressions videos), and 2) to present a classification system that sets a challenging comparison performance on this dataset. The iCOPEvid dataset contains 234 videos of 49 neonates experiencing a set of noxious stimuli, a period of rest, and an acute pain stimulus. From these videos 20 s segments are extracted and grouped into two classes: pain (49) and nopain (185), with the nopain video segments handpicked to produce a highly challenging dataset. An ensemble of twelve global and local descriptors with a Bag-of-Features approach is utilized to improve the performance of some new descriptors based on Gaussian of Local Descriptors (GOLD). The basic classifier used in the ensembles is the Support Vector Machine, and decisions are combined by sum rule. These results are compared with standard methods, some deep learning approaches, and 185 human assessments. Our best machine learning methods are shown to outperform the human judges.

In August 2004, the Library Collections and Systems team at Lehigh University released MyLibrary @ Lehigh within the campus portal to the university community. The purpose of this article is to explain how what began as an integration strategy of the library's electronic resources into one complete stand‐alone application became the library's response to the fast‐growing campus portal.

Explains how MyLibrary@Lehigh was developed and implemented.

It became evident during its development and integration stages that MyLibrary@Lehigh would greatly enhance the success and usage of the campus portal. As a repository of all of the library's electronic resources, MyLibrary@Lehigh has become the one‐stop shop for the library within the campus‐wide, one‐stop shop of the campus portal.

MyLibrary has become a clear choice as an open‐source solution.

Bankruptcy prediction has attracted a great deal of research in the data mining/machine learning community, due to its significance in the world of accounting, finance, and investment. This chapter examines the influence of different dimension reduction techniques on decision tree model applied to the bankruptcy prediction problem. The studied techniques are principal component analysis (PCA), sliced inversed regression (SIR), sliced average variance estimation (SAVE), and factor analysis (FA). To focus on the impact of the dimension reduction techniques, we chose only to use decision tree as our predictive model and “undersampling” as the solution to the issue of data imbalance. Our computation shows that the choice of dimension reduction technique greatly affects the performances of predictive models and that one could use dimension reduction techniques to improve the predictive power of the decision tree model. Also, in this study, we propose a method to estimate the true dimension of the data.

This chapter is a contribution to a recent restricted literature dealing with the return of microfinance investment in the financial markets. We study the performance of public microfinance investment vehicles (MIVs). Microfinance is an asset class with a double bottom line: social and financial returns have to be generated. Despite a significant currency risk, we find that the integration of microfinance assets diversifies the investor’s risks and improves the efficient frontier. We conclude that microfinance institutions, via investment vehicles, are likely to attract capital from socially responsible investors seeking new investment opportunities.

Purpose: The purpose of this chapter is to review and critically evaluate robots, artificial intelligence and service automation (RAISA) applications in the restaurant industry to educate professors, graduate students, and industry professionals.

Design/methodology/approach: This chapter is a survey of applications of RAISA in restaurants. The chapter is based on the review of professional and peer-reviewed academic literature, and the industry insight section was prepared based on a 50-minute interview with Mr. Juan Higueros, Chief Operations Officer of Bear Robotics.

Findings: Various case studies presented in this chapter illustrate numerous possibilities for automation: from automating a specific function to complete automation of the front of the house (e.g., Eatsa) or back of the house (e.g., Spyce robotic kitchen). The restaurant industry has already adopted chatbots; voice-activated and biometric technologies; robots as hosts, food runners, chefs, and bartenders; tableside ordering; conveyors; and robotic food delivery.

Practical implications: The chapter presents professors and students with a detailed overview of RAISA in the restaurant industry that will be useful for educational and research purposes. Restaurant owners and managers may also benefit from reading this chapter as they will learn about the current state of technology and opportunities for RAISA implementation.

Originality/value: To the best of the authors’ knowledge, this chapter presents the first systematic and in-depth review of RAISA technologies in the restaurant industry.

The growth of the global labor force and business analytics has significantly impacted human resource management (HRM). Human resource (HR) analytics is an emerging field that creates value for employees and organizations. By examining the existing studies on HR analytics, the paper systematically reviews the literature to identify active research areas and establish a roadmap for future studies in HR analytics.

A portfolio of 503 articles collected from the Scopus database was reviewed. The study has adopted a Latent Dirichlet allocation (LDA) topic modeling approach to identify significant themes in the literature.

The HR analytics research domain is classified into four categories: HR functions, statistical techniques, organizational outcomes and employee characteristics. The study has also developed a framework for organizations adopting HR analytics. Linking HR with blockchain technology, explainable artificial intelligence and Metaverse are the areas identified for future researchers.

The framework will assist practitioners in identifying statistical techniques for optimizing various HR functions. The paper discovers that by implementing HR analytics, HR managers and business partners can run reports, make dashboards and visualizations and make evidence-based decision-making.

The previous studies have not applied any machine learning techniques to identify the topics in the extant literature. The paper has applied machine learning tools, making the review more robust and providing an exhaustive understanding of the domain.

Let G=(V,E) be a graph. The complement of G is the graph G¯:=(V,[V]2 \ E) where [V]2 is the set of pairs {x,y} of distinct elements of V. If K is a subset of V, the restriction of G to K is the graph G↾K:=(K,[K]2∩E). We prove that if G=(V,E) is a graph and k is an integer, 2 ≤ k ≤ v−2, then there is a k -element subset K of V such that e(G¯↾K)≠e(G↾K), moreover the condition k ≤ v−2 is optimal. We also study the case e(G¯↾K) ≢ e(G↾K)(mod p) where p is a prime number. Following a question from M.Pouzet, we show this: Let G=(V,E) be a graph with v vertices. If e(G) ≠ e(G¯) (resp. e(G)=e(G¯)) then there is an increasing family (Hn)2≤n≤v (resp. (Hn)2≤n≤v−2) of n -element subsets Hn of V such that e(G↾Hn)≠ e(G¯↾Hn) for all n. Similarly if e(G) ≢ e(G¯) (mod p) where p is a prime number, p>v−2, then there is an increasing family (Hn)2≤n≤v of n -element subsets Hn of V such that e(G↾Hn) ≢ e(G¯↾Hn)(mod p) for all integer n∈{2,3,…,v}.