A dynamic scorecard for monitoring baseline performance with application to tracking a mortgage portfolio

A principled technique for monitoring the performance of a consumer credit scorecard through time is derived from Kalman filtering. Standard approaches sporadically compare certain characteristics of the new applicants with those predicted from the scorecard. The new approach systematically updates the scorecard combining new applicant information with the previous best estimate. The dynamically updated scorecard is tracked through time and compared to limits calculated by sequential simulation from the baseline scorecard. The observation equation of the Kalman filter is tailored to take the results of fitting local scorecards by logistic regression to batches of new clients that arrive in the current time interval. The states in the Kalman filter represent the true or underlying score for each attribute in the card: the parameters of the logistic regression. Their progress in time is modelled by a random walk and the filter provides the best estimate of the scores using past and present information. We illustrate the technique using a commercial mortgage portfolio and the results indicate significant emerging deficiencies in the baseline scorecard.     

Local search algorithms for the min-max loop layout problem

In the min-max loop layout problem, machines are to be arranged around a loop of conveyor belt. The ordering of the machines dictates the number of circuits of the conveyor belt required to manufacture each of several products. The goal is to find an ordering of the machines that minimises the maximum number of circuits required for the manufacture of any of the products. Since the problem is strongly NP-hard, the study of heuristic methods is of interest. This paper proposes iterated descent and tabu search algorithms, and a randomised insertion algorithm. Results of extensive computational tests show that all of our algorithms outperform a previously known algorithm that applies a greedy heuristic to the solution of a linear programming relaxation. The best quality solutions are obtained with iterated descent. This adds further evidence to the belief that iterated descent can produce high quality solutions to a variety of combinatorial optimisation problems. Moreover, unlike some other local search algorithms, iterated descent does not require much tuning in order to be competitive.     

Atomic Xα calculations based on EHFS(α) = Eexp

The total energies and one-electron energies for first- and second-row atoms were calculated by using the Hartree–Fock and the Hartree–Fock-Slater Hamiltonian with X_α orbitals, u_i(α_exp); α was parametrized from E^HFS (α_exp) = E^exp. The E^HF (α_exp) total energies are always higher than the Hartree–Fock energies for the atoms. The relation of the calculated ionization potential to the experimental ionization potential depends on the α used to define u_i(α), α_exp, or α_HF.     

Ionization potentials,electron affinities,electronegativities, and hardnesses of fractional charged atoms from the density-functional theory

The electron-correlation and self-interaction corrected generalized exchange local-spin-density functional theory with the Gopinathan, Whitehead, and Bogdanovic Fermi-hole parameters has been employed to give self-consistent field calculations for the quark atoms, the first- and second-order positive ions, and the first- and second-order negative ions of the quark atoms with fractional nuclear charges $ Z = N \pm \frac{1}{3} $ and $ Z = N \pm \frac{2}{3} $. A special technique to obtain the converged second-order negative ions is discussed. The first and second ionization potentials and electron affinities are calculated by the differences of the total energies between the ionized and nonionized systems and compared with the empirical inter-extrapolation results. The agreement between the present calculations and the inter-extrapolated results is excellent for the ionization potentials and reasonably good for the electron affinities of the quark atoms. Finally, the calculated ionization potentials and electron affinities are used in obtaining the electronegativities and hardnesses for these quark atoms.     

Shaping the density to fit one-electron properties: Constrained RHF calculations on N2, FH,CO, and LiH

The molecular density required to give the correct values for one-electron properties is rarely given by wave functions obtained from variation methods based on the total energy or the eigenvalues. Perhaps if we knew how the density should be shaped in any particular volume to fit a particular property, the whole molecular density might then be properly described to fit the whole volume. The secant-parametrization procedure is used to constrain minimum basis set RHF wave functions for N₂, FH, CO, and LiH to determine the effects of different constraints on RHF wave functions, and to see how constraints improve the quality of small basis set RHF wave functions. One-electron property expectation values, energies, and unweighted and property weighted populations and electron density difference profiles are used to analyze the constrained wavefunctions. With the information from the constrained wave functions it should be possible to select a LCAO -CI basis and states to give the correct density for all properties. This should map onto the constrained wave function in the region of the constraint and at the same time minimize the energy of the total molecular wave function. Such a density would be suitable for the density analyses favored by Bader and Nguyen-Dang [Adv. Quantum Chem. 14 , 113 (1981)], Mezey [Theor. Chim. Acta 54 , 95 (1980); 58 , 309 (1981); 59 , 321 (1981)], and March [Theoretical Chemistry (Royal Society of Chemistry, London, 1981), Vol. 4, p. 158], and show the real atom needed to generate the molecule.     

Non self-consistent field theory

Roothaan's SCF method [2] is reformulated so that two non-SCP methods are developed to solve the eigenequation. The results from these methods can be used as starting eigenfunctions for Roothaan's SCF method.

---

Zusammenfassung Die self-consistent field Methode von Roothaan [2] wird neu formuliert: dabei werden zwei Nicht-SCF Methoden zur Lösung der Eigenwertgleichung entwickelt. Deren Resultate können als Anfangsfunktionen bei Rechnungen mit der SCF-Methode von Roothaan benutzt werden.

---

Résumé La méthode SCF de Roothaan [2] est reformulée en développant deux méthodes non-SCF pour la solution de l'équation aux valeurs propres. Les résultats de ces deux méthodes peuvent être utilisées comme fonctions de départ pour la méthode SCF de Roothaan.

---

---

The title Theoretical Chemistry has been transferred to the Division of Physical Chemistry.     

Transferable atom equivalent multicentered multipole expansion method

The transferability of atomic and functional group properties is an implicit concept in chemistry. The work presented here describes the use of Transferable Atom Equivalents (TAE) to represent molecular electrostatic potential fields through the use of integrated atomic multipole moments that are associated with each TAE atom type used in the reconstruction. TAE molecular surface distributions of electrostatic potentials are compared with analytical ab initio and empirical (Gasteiger) partial charge reference models for several conformations of test peptides. Surface electrostatic potential distributions computed using TAE multipole representations were found to converge at the octopole level, with incremental improvement observed when hexadecapoles were included. Molecular electrostatic potential fields that were produced using the TAE method were observed to be responsive to conformational changes and to compare well with ab initio reference distributions. Generation of TAE atom types and their associated multipoles does not involve fitting to sample electrostatic potential fields, but rather utilizes integrated AIM atomic electron density distributions within representative chemical environments. The RECON program was used for TAE reconstruction. RECON is capable of processing 5,000 drug-sized molecules or 25 proteins per minute per 1.7 GHz P4 Linux processor.     

A simple atmospheric microwave-excited emissive detector for gas chromatography

An atmospheric pressure microwave-excited emissive argon plasma is evaluated for use as a detector in conventional gas-chromatographic analysis. A range of carbon-, oxygen-, nitrogen- and halogen-containing compounds was investigated in order to optimize the sensitivity, linear working range and plasma operating conditions. For the non-selective determination of carbon-containing compounds, the atomic carbon emission at 247.9 nm was found to be most useful. At this wavelength the sensitivities for all the compounds investigated were directly proportional to the amount of carbon in the compounds. The limit of detection for these compounds at 247.9 nm was 1.94·10^-l0 g of carbon per sec. The atomic lines at 206.2 nm (I) and 470.5 nm (Br) were the best for the selective determination of iodine- and bromine-containing compounds respectively, and a new band emission at 256 nm, characteristic of chlorinecontaining compounds, gave good sensitivity.