WHO WE ARE / AWARDS
The business challenge was to be able to identify patients in a hospital, with high risk of contracting nosocomial pneumonia during a surgical procedure.The LatentView team developed a predictive model which would score patients on the risk of contracting nosocomial pneumonia. After evaluating different techniques, Logistic regression was the chosen technique to develop the predictive model, which had an AUC of 0.8304.Optimization algorithms, Bayesian networks, CART, etc were explored over and above the usual predictive modeling techniques.
The Institute for Operations Research and the Management Sciences (INFORMS) is the largest professional society in the world for professionals in the field of operations research (O.R.). It was established in 1995 with the merger of the Operations Research Society of America (ORSA) and The Institute of Management Sciences (TIMS).
LatentView among Top Performers at PAKDD 2009 Data Mining Competition
The business problem was to develop a predictive risk assessment model to rank order credit card applicants on their probability to default on payments. LatentView developed a suite of models to derive the final score for every customer and rank-order them on their default propensity.LatentView used three techniques namely – Regression, Decision Tree & Stochastic Boosting Algorithm. The differentiator was in the innovative application of the different techniques to develop the solution. The approach involved developing an ensemble of models using the techniques on various random samples and using novel methods to combine the individual predictions to determine the final prediction.
The 13th Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD-09) is a major international conference in the areas of data mining and knowledge discovery. It provides an international forum for researchers and industry practitioners to share their new ideas, original research results and practical development experiences from all KDD-related areas including data mining, data warehousing, machine learning, databases, statistics, knowledge acquisition and automatic scientific discovery, data visualization, causal induction and knowledge-based systems.
LatentView finishes 4th in the Fast track of the KDD 2009 Data Mining Contest
The challenge was to build three solutions that would predict the propensity of the customer to buy another product (Appetency), buy a product of higher value (Up-sell) and switch provider (Churn) in the telecommunications context. Key feature of the challenge was to develop a fast scoring solution for the three business problems on a large database (15,000 features) within five days from the release of the challenge data. The challenge was to beat the in-house system developed by Orange Labs. It was an opportunity to prove our abilities to deal with a large database, including heterogeneous noisy data (numerical and categorical variables), and unbalanced class distributions.
Over the course of the challenge, we developed around 160 models using one or more techniques. Our over-all approach in developing these models and our final solution can be summarized as “Ensemble of Models developed using suite of Logistic regression models, Gradient Boosting, Adaptive Logistic Regression, Decision Tree and Naive Bayes algorithm on various random samples and using innovative techniques to combine the predictions from various models to develop the final score.”
The annual ACM SIGKDD conference is the premier international forum for data mining researchers and practitioners from academia, industry, and government to share their ideas, research results and experiences.
Accolades in the recently concluded data mining competitions are a testimony to LatentView’s expertise in data mining and predictive analytics.LatentView participated in three of the major contests in the last one year and has secured top positions in all the three contests.
LatentView adjudged as one of the winners at the INFORMS Data Mining Contest 2008The Institute for Operations Research and the Management Sciences (INFORMS) is the largest professional society in the world for professionals in the field of operations research (O.R.). It was established in 1995 with the merger of the Operations Research Society of America (ORSA) and The Institute of Management Sciences (TIMS).
LatentView among Top Performers at PAKDD 2009 Data Mining Competition
The business problem was to develop a predictive risk assessment model to rank order credit card applicants on their probability to default on payments. LatentView developed a suite of models to derive the final score for every customer and rank-order them on their default propensity.LatentView used three techniques namely – Regression, Decision Tree & Stochastic Boosting Algorithm. The differentiator was in the innovative application of the different techniques to develop the solution. The approach involved developing an ensemble of models using the techniques on various random samples and using novel methods to combine the individual predictions to determine the final prediction.
The 13th Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD-09) is a major international conference in the areas of data mining and knowledge discovery. It provides an international forum for researchers and industry practitioners to share their new ideas, original research results and practical development experiences from all KDD-related areas including data mining, data warehousing, machine learning, databases, statistics, knowledge acquisition and automatic scientific discovery, data visualization, causal induction and knowledge-based systems.
LatentView finishes 4th in the Fast track of the KDD 2009 Data Mining Contest
Over the course of the challenge, we developed around 160 models using one or more techniques. Our over-all approach in developing these models and our final solution can be summarized as “Ensemble of Models developed using suite of Logistic regression models, Gradient Boosting, Adaptive Logistic Regression, Decision Tree and Naive Bayes algorithm on various random samples and using innovative techniques to combine the predictions from various models to develop the final score.”
The annual ACM SIGKDD conference is the premier international forum for data mining researchers and practitioners from academia, industry, and government to share their ideas, research results and experiences.
