No description available.

No description available.

Made available in DSpace on 2014-06-12T15:05:57Z (GMT). No. of bitstreams: 2 arquivo1182_1.pdf: 1252336 bytes, checksum: f525fed8d87638ae39fd882853bce5ca (MD5) license.txt: 1748 bytes, checksum: 8a4605be74aa9ea9d79846c1fba20a33 (MD5) Previous issue date: 2007 / Conselho Nacional de Desenvolvimento Científico e Tecnológico / Esta dissertação propôs o estudo da viabilidade da utilização de modelos de credit scoring em uma instituição educacional de ensino médio e fundamental, atuante na rede privada na cidade do Recife PE. A utilização deste tipo de modelagem é bastante difundida em instituições financeiras, no entanto sua prática no segmento de serviços apresenta-se em estágio embrionário, onde seus benefícios ainda são pouco conhecidos. A utilização de modelos como mecanismos de apoio ao gerenciamento de serviços educacionais assume importante relevância, pois este segmento tem sido severamente penalizado com elevados índices de inadimplência, aliado a uma legislação desfavorável quanto a cobrança de débitos vencidos. No desenvolvimento dos modelos de credit scoring foram utilizados as técnicas de análise discriminante, regressão logística e rede neural artificial, onde sua viabilidade foi avaliada ao se comparar a performance da previsão dos modelos com o percentual de acertos obtido pelo critério de chances. Os resultados demonstram que a análise discriminante obteve o melhor desempenho na previsão do grupo dos inadimplentes, com 80% de acerto. Por outro lado, os modelos baseados na regressão logística e rede neural artificial alcançaram o mais alto nível de acerto no grupo dos adimplentes, ambos com 93,48%. Sendo assim a modelagem de credit scoring apresentou-se como um instrumento de gestão de risco viável para a instituição de educação pesquisada

Ingeniera Civil Industrial / En Chile, durante los últimos años, ha habido un aumento en la tasa de evasión de impuestos, observándose que, en particular para el IVA esta cifra ha crecido un 8,5% con respecto el año 2007. Cada punto porcentual se traduce en una pérdida de recaudación de 350 millones de dólares, por lo que se hace necesario el diseño de un plan que revierta este efecto. La presente memoria consiste en el desarrollo de un modelo estadístico que permita predecir el default en la declaración y pago de IVA para los contribuyentes de Micro y Pequeña empresa. Teniendo como resultado la detección de las variables que más influyen en esta conducta, la probabilidad de default de cada contribuyente para el período tributario julio 2013 y la identificación del porcentaje de default según zona geográfica. La metodología de trabajo contempla las etapas de entendimiento del negocio, extracción de la información, preparación de datos, modelamiento y finalmente, la interpretación de los resultados. Dada la inexistencia de trabajos publicados en materia de default tributario, el sustento teórico de esta investigación se basa en el credit scoring, técnica utilizada en la industria bancaria. Se toma como fuente de información el Data Warehouse del SII, con el que se construyen 55 variables que reflejan características demográficas, índices de liquidez, movimientos de caja, comportamiento y tamaño del negocio. Se prueban tres algoritmos de clasificación: árboles de decisión, regresión logística y redes neuronales. Se elige como modelo definitivo el de regresión logística, dada su clara interpretabilidad y buena capacidad de predicción, que alcanza una precisión global de 68,81%, especificidad de 67,29% y sensibilidad de 68,88%, cumpliendo con el objetivo inicial de obtener el mejor modelo predictivo balanceado posible. Respecto de las variables, los resultados indican que las de mayor relevancia se relacionan con el historial de cumplimiento del F29, dónde se presenta una clara segmentación de los contribuyentes en tres tipos de conducta. Respecto de la identificación por zona geográfica, se aprecia el mayor porcentaje de default para la zona del norte grande del país, y el menor para la Región de Magallanes y la zona Centro y Oriente de la Región Metropolitana. Se concluye que la presencia de errores en la información proporcionada por los contribuyentes, así como la cantidad de campos nulos encontrados, hace que aún no se cuente con una base de datos óptima para aplicar la técnica de credit scoring. El SII se encuentra trabajando en esta línea, lo que permitirá en un futuro obtener mejores resultados. Como recomendación final se propone utilizar las predicciones obtenidas para diseñar un plan de medidas preventivas, así como también evaluar el desarrollo de un modelo de alta precisión y baja especificidad que se enfoque en la detección de los defaulters más críticos.

As a low fidelity work simulation, Situational Judgment Tests (SJTs) are an affordable and practical way of empirically linking training and on-the-job performance, thereby providing a viable means of evaluating training effectiveness. An issue, when utilizing SJTs, is deciding the appropriate manner in which the SJT should be scored. Traditional SJT scoring methodologies, while successfully utilized for selection and prediction, pose specific challenges when applied to a SJT designed to evaluate the effectiveness of a training program. This study discusses the shortcomings of traditional SJT scoring methodologies when used in the evaluation context. To overcome these challenges, an innovative scoring methodology, the Hybrid methodology, is presented. This study provides the detailed description of the Hybrid scoring key creation, and compares the Hybrid scoring key with two traditional scoring keys (Subject Matter Expert (SME) and Respondent-based scoring keys). Responses from a military training program are utilized to illustrate the distinctive effects of using the three different scoring approaches. The superiority of the hybrid scoring key, due to increased confidence in the key's accuracy, and findings regarding training evaluation are discussed. Future research directions and practical applications of the research are also discussed. / Master of Science

Thesis advisor: Matthias von Davier / Automated scoring has received considerable attention in educational measurement, even before the era of artificial intelligence. However, its application to constructed response (CR) items in international large-scale assessments (ILSAs) remains largely underexplored due to the complexity of tackling multilingual responses spanning often over 100 different language versions. This doctoral dissertation aims to address this issue by progressively expanding the scope of automated scoring from several countries in TIMSS 2019 to all participating countries in TIMSS 2023. We delved into the feasibility of automated scoring across diverse linguistic landscapes, encompassing high-resource and low-resource languages. We examined two machine learning methodologies—supervised and unsupervised learning—integrating them with cutting-edge machine translation techniques. Our findings demonstrated that automated scoring can serve as a reliable and cost-effective measure for quality assurance in ILSAs, significantly reducing the reliance on secondary human raters. Ultimately, the adoption of automated scoring instead of human scoring in the foreseeable future will promote the broader use of innovative open-item formats in ILSAs. / Thesis (PhD) — Boston College, 2024. / Submitted to: Boston College. Lynch School of Education. / Discipline: Measurement, Evaluation, Statistics & Assessment.

published_or_final_version / Education / Master / Master of Education

The promise of increased student achievement through educational reform is delivered still-born if teachers do not know how to implement complex instructional practices and sophisticated analysis of student performance. Metacognitive awareness is crucial to the adoption and application of proven educational initiatives. Teachers who successfully implement criterion-referenced assessment instruction, scoring rubrics, transfer to their students the metacognitive knowledge and skills of how to learn. This study is predicated on the research assumptions that metacognition and its attendant skills are critical to the successful implementation of scoring rubrics. A researcher-developed instrument, the Scoring Rubrics Inventory (SRI) and the Metacognitive Awareness Inventory (MAI) were distributed to core-subject teachers from three large public schools in Southwest Louisiana. From a population of sixty-eight (N=68) voluntary participants, eighteen teacher-participants self-reported as high implementers of scoring rubrics, thirty-nine as mid-level implementers and eleven as low-level implementers. From this population, twelve subjects were randomly selected (four high, four mid-level, and four lowlevel implementers) by an outside rater for double-blind observations and interviews. Pearson Product Moment correlations of the SRI and the MAI revealed five significant pairings using an alpha level of .05. The statistical results, coupled with the observation and interview findings from the sample-subjects established the consistency and stability of the Scoring Rubrics Inventory. Further, the totality of the results reported here support the research hypothesis of the study: H1: There is a significant correlation between the metacognitive awareness of secondary school core-subject teachers and the successful implementation of scoring rubrics. The results of the study indicated that secondary school core-subject teachers who successfully implement scoring rubrics possess a metacognitive awareness that transcends professional development training. The findings also suggested that teacher-participants who do not implement scoring rubrics either cannot or lack commitment to the innovation. Implications for teacher educators and school leaders indicated the need to: identify those persons who require additional professional development training; include operational strategies and modeling of successful implementation during training; and maintain a consistent training program in scoring rubrics. Recommendations for future research were offered.

Dissertation thesis deals with the topic of capital adequacy of financial institutions and tries to solve the problem of default and its prediction. In the theoretical part, the thesis provides summarization of historic and current approaches to capital adequacy of financial institutions and also presents currently used methodology of scoring models, which predict default of companies. Application part of the thesis aims to develop a scoring model, which would be usable by financial institutions for evaluation of their clients and their tendency to default.

Os modelos de credit scoring têm sido bastante difundidos nos últimos anos como uma importante ferramenta para agilizar e tornar mais confiável o processo de concessão de crédito por parte das instituições financeiras. Esses modelos são utilizados para classificar os clientes em relação a seus riscos de inadimplência. Neste trabalho, é avaliada a aplicabilidade de uma nova metodologia, baseada em pseudo-valores, como alternativa para a construção de modelos de credit scoring. O objetivo é compará-la com abordagens tradicionais como a regressão logística e o modelo de riscos proporcionais de Cox. A aplicação prática é feita para dados de operações de crédito pessoal sem consignação, coletados do Sistema de Informações de Crédito do Banco Central do Brasil. As performances dos modelos são comparadas utilizando a estatística de Kolmogorov-Smirnov e a área sob a curva ROC. / Credit Scoring models have become popular in recent years as an important tool in the credit granting process, making it more expedite and reliable. The models are mainly considered to classify customers according to their default risk. In this work we evaluate the apllicability of a new methodology, based on pseudo-values, as an alternative to constructing credit scoring models. The objective is to compare this novel methodology with traditional approaches such as logistic regression and Cox proportional hazards model. The models are applied to a dataset on personal credit data, collected from the Credit Information System of Central Bank of Brazil. The performances of the models are compared via Kolmogorov-Smirnov statistic and the area under ROC curve.