Reframing reverse logistics, reducing costs

A Work Project, presented as part of the requirements for the Award of a Masters Degree in Management from the NOVA – School of Business and Economics

Comparison of three expert elicitation methods for logistic regression on predicting the presence of the threatened brush-tailed rock-wallaby petrogale penicillata

Numerous expert elicitation methods have been suggested for generalised linear models (GLMs). This paper compares three relatively new approaches to eliciting expert knowledge in a form suitable for Bayesian logistic regression. These methods were trialled on two experts in order to model the habitat suitability of the threatened Australian brush-tailed rock-wallaby (Petrogale penicillata). The first elicitation approach is a geographically assisted indirect predictive method with a geographic information system (GIS) interface. The second approach is a predictive indirect method which uses an interactive graphical tool. The third method uses a questionnaire to elicit expert knowledge directly about the impact of a habitat variable on the response. Two variables (slope and aspect) are used to examine prior and posterior distributions of the three methods. The results indicate that there are some similarities and dissimilarities between the expert informed priors of the two experts formulated from the different approaches. The choice of elicitation method depends on the statistical knowledge of the expert, their mapping skills, time constraints, accessibility to experts and funding available. This trial reveals that expert knowledge can be important when modelling rare event data, such as threatened species, because experts can provide additional information that may not be represented in the dataset. However care must be taken with the way in which this information is elicited and formulated.

Modeling crash outcome probabilities at rural intersections: Application of hierarchical binomial logistic models

It is important to examine the nature of the relationships between roadway, environmental, and traffic factors and motor vehicle crashes, with the aim to improve the collective understanding of causal mechanisms involved in crashes and to better predict their occurrence. Statistical models of motor vehicle crashes are one path of inquiry often used to gain these initial insights. Recent efforts have focused on the estimation of negative binomial and Poisson regression models (and related deviants) due to their relatively good fit to crash data. Of course analysts constantly seek methods that offer greater consistency with the data generating mechanism (motor vehicle crashes in this case), provide better statistical fit, and provide insight into data structure that was previously unavailable. One such opportunity exists with some types of crash data, in particular crash-level data that are collected across roadway segments, intersections, etc. It is argued in this paper that some crash data possess hierarchical structure that has not routinely been exploited. This paper describes the application of binomial multilevel models of crash types using 548 motor vehicle crashes collected from 91 two-lane rural intersections in the state of Georgia. Crash prediction models are estimated for angle, rear-end, and sideswipe (both same direction and opposite direction) crashes. The contributions of the paper are the realization of hierarchical data structure and the application of a theoretically appealing and suitable analysis approach for multilevel data, yielding insights into intersection-related crashes by crash type.

A hybrid approach to combining CART and logistic regression for stock ranking

The benefits of applying tree-based methods to the purpose of modelling financial assets as opposed to linear factor analysis are increasingly being understood by market practitioners. Tree-based models such as CART (classification and regression trees) are particularly well suited to analysing stock market data which is noisy and often contains non-linear relationships and high-order interactions. CART was originally developed in the 1980s by medical researchers disheartened by the stringent assumptions applied by traditional regression analysis (Brieman et al. [1984]). In the intervening years, CART has been successfully applied to many areas of finance such as the classification of financial distress of firms (see Frydman, Altman and Kao [1985]), asset allocation (see Sorensen, Mezrich and Miller [1996]), equity style timing (see Kao and Shumaker [1999]) and stock selection (see Sorensen, Miller and Ooi [2000])...

Assessment and rationalization of genetic diversity of Papua New Guinea taro (Colocasia esculenta) using SSR DNA fingerprinting

Taro (Colocasia esculenta) accessions were collected from 15 provinces of Papua New Guinea (PNG). The collection, totalling 859 accessions was collated for characterization and a core collection of 81 accessions (10%) was established on the basis of characterization data generated on 30 agro-morphological descriptors, and DNA fingerprinting using seven SSR primers. The selection of accessions was based on cluster analysis of the morphological data enabling initial selection of 20% accessions. The 20% sample was then reduced and rationalized to 10% based on molecular data generated by SSR primers. This represents the first national core collection of any species established in PNG based on molecular markers. The core has been integrated with core from other Pacific Island countries, contributing to a Pacific regional core collection, which is conserved in vitro in the South Pacific Regional Germplasm Centre at Fiji. The core collection is a valuable resource for food security of the South Pacific region and is currently being utilized by the breeding programmes of small Pacific Island countries to broaden the genetic base of the crop.

Fluorescent Fields: electric lighting and the rationalization of the modern corporate workplace

This study investigates the implications of the introduction of electric lighting systems, building technologies, and theories of worker efficiency on the deep spatial and environmental transformations that occurred within the corporate workplace during the twentieth century. Examining the shift from daylighting strategies to largely artificially lit workplace environments, this paper argues that electric lighting significantly contributed to the architectural rationalization of both office work and the modern office environment. Contesting the historical and critical marginalization of lighting within the discourse of the modern built environment, this study calls for a reassessment of the role of artificial lighting in the development of the modern corporate workplace. Keywords: daylighting, fluorescent lighting, rationalization, workplace design

A fast dual algorithm for kernel logistic regression

This paper gives a new iterative algorithm for kernel logistic regression. It is based on the solution of a dual problem using ideas similar to those of the Sequential Minimal Optimization algorithm for Support Vector Machines. Asymptotic convergence of the algorithm is proved. Computational experiments show that the algorithm is robust and fast. The algorithmic ideas can also be used to give a fast dual algorithm for solving the optimization problem arising in the inner loop of Gaussian Process classifiers.

On the bifurcation in a “modulated” logistic map

We comment on a paper by Luang [On the bifurcation in a ''modulated'' logistic map, Physics Letters A 194(1994) 57]. The numerical evidence given in that paper, for a peculiar type of bifurcation, is shown to be incorrect. The causes of such anomalous results are explained. An accurate bifurcation diagram for the map is also given.

Learning polyhedral classifiers using logistic function

In this paper we propose a new algorithm for learning polyhedral classifiers. In contrast to existing methods for learning polyhedral classifier which solve a constrained optimization problem, our method solves an unconstrained optimization problem. Our method is based on a logistic function based model for the posterior probability function. We propose an alternating optimization algorithm, namely, SPLA1 (Single Polyhedral Learning Algorithm1) which maximizes the loglikelihood of the training data to learn the parameters. We also extend our method to make it independent of any user specified parameter (e.g., number of hyperplanes required to form a polyhedral set) in SPLA2. We show the effectiveness of our approach with experiments on various synthetic and real world datasets and compare our approach with a standard decision tree method (OC1) and a constrained optimization based method for learning polyhedral sets.

Rationalization and prediction of drug resistant mutations in targets for clinical anti-tubercular drugs

Resistance to therapy limits the effectiveness of drug treatment in many diseases. Drug resistance can be considered as a successful outcome of the bacterial struggle to survive in the hostile environment of a drug-exposed cell. An important mechanism by which bacteria acquire drug resistance is through mutations in the drug target. Drug resistant strains (multi-drug resistant and extensively drug resistant) of Mycobacterium tuberculosis are being identified at alarming rates, increasing the global burden of tuberculosis. An understanding of the nature of mutations in different drug targets and how they achieve resistance is therefore important. An objective of this study is to first decipher sequence as well as structural bases for the observed resistance in known drug resistant mutants and then to predict positions in each target that are more prone to acquiring drug resistant mutations. A curated database containing hundreds of mutations in the 38 drug targets of nine major clinical drugs, associated with resistance is studied here. Mutations have been classified into those that occur in the binding site itself, those that occur in residues interacting with the binding site and those that occur in outer zones. Structural models of the wild type and mutant forms of the target proteins have been analysed to seek explanations for reduction in drug binding. Stability analysis of an entire array of 19 mutations at each of the residues for each target has been computed using structural models. Conservation indices of individual residues, binding sites and whole proteins are computed based on sequence conservation analysis of the target proteins. The analyses lead to insights about which positions in the polypeptide chain have a higher propensity to acquire drug resistant mutations. Thus critical insights can be obtained about the effect of mutations on drug binding, in terms of which amino acid positions and therefore which interactions should not be heavily relied upon, which in turn can be translated into guidelines for modifying the existing drugs as well as for designing new drugs. The methodology can serve as a general framework to study drug resistant mutants in other micro-organisms as well.

Large-Scale Elastic Net Regularized Linear Classification SVMs and Logistic Regression

Elastic Net Regularizers have shown much promise in designing sparse classifiers for linear classification. In this work, we propose an alternating optimization approach to solve the dual problems of elastic net regularized linear classification Support Vector Machines (SVMs) and logistic regression (LR). One of the sub-problems turns out to be a simple projection. The other sub-problem can be solved using dual coordinate descent methods developed for non-sparse L2-regularized linear SVMs and LR, without altering their iteration complexity and convergence properties. Experiments on very large datasets indicate that the proposed dual coordinate descent - projection (DCD-P) methods are fast and achieve comparable generalization performance after the first pass through the data, with extremely sparse models.

Structural analysis of dihydrofolate reductases enables rationalization of antifolate binding affinities and suggests repurposing possibilities

Antifolates are competitive inhibitors of dihydrofolate reductase ( DHFR), a conserved enzyme that is central to metabolism and widely targeted in pathogenic diseases, cancer and autoimmune disorders. Although most clinically used antifolates are known to be target specific, some display a fair degree of cross-reactivity with DHFRs from other species. A method that enables identification of determinants of affinity and specificity in target DHFRs from different species and provides guidelines for the design of antifolates is currently lacking. To address this, we first captured the potential druggable space of a DHFR in a substructure called the `supersite' and classified supersites of DHFRs from 56 species into 16 `site-types' based on pairwise structural similarity. Analysis of supersites across these site-types revealed that DHFRs exhibit varying extents of dissimilarity at structurally equivalent positions in and around the binding site. We were able to explain the pattern of affinities towards chemically diverse antifolates exhibited by DHFRs of different site-types based on these structural differences. We then generated an antifolate-DHFR network by mapping known high-affinity antifolates to their respective supersites and used this to identify antifolates that can be repurposed based on similarity between supersites or antifolates. Thus, we identified 177 human-specific and 458 pathogen-specific antifolates, a large number of which are supported by available experimental data. Thus, in the light of the clinical importance of DHFR, we present a novel approach to identifying differences in the druggable space of DHFRs that can be utilized for rational design of antifolates.

Temporally adaptive estimation of logistic classifiers on data streams

Modern technology has allowed real-time data collection in a variety of domains, ranging from environmental monitoring to healthcare. Consequently, there is a growing need for algorithms capable of performing inferential tasks in an online manner, continuously revising their estimates to reflect the current status of the underlying process. In particular, we are interested in constructing online and temporally adaptive classifiers capable of handling the possibly drifting decision boundaries arising in streaming environments. We first make a quadratic approximation to the log-likelihood that yields a recursive algorithm for fitting logistic regression online. We then suggest a novel way of equipping this framework with self-tuning forgetting factors. The resulting scheme is capable of tracking changes in the underlying probability distribution, adapting the decision boundary appropriately and hence maintaining high classification accuracy in dynamic or unstable environments. We demonstrate the scheme's effectiveness in both real and simulated streaming environments. © Springer-Verlag 2009.