Combining two-stage decomposition based machine learning methods for annual runoff forecasting

AbstractIdentifying investment patterns as part of customer segmentation is one of the most important tasks in retail banking. Clustering customers effectively is an important element of improving marketing policy and strategic planning. There are several methods for identifying similar groups of customers and describing their characteristics to offer them appropriate products. However, using machine learning methods is rare, and the application is limited for certain types of data. The aim of this study is to investigate the benefits of using a two-stage clustering method using neural-network-based Kohonen self-organizing maps followed by hierarchical clustering for identifying the investment patterns of potential retail banking customers. The unique benefit of this method is the ability to use both categorical and numerical variables at the same time. This research examined 1,542 responses received for an online investment survey, focusing on the questions that are related to the respondents’ investment preferences and their current financial assets. The research utilizes descriptive statistics and multiple correspondence analysis (MCA) to understand the variables and Kohonen self-organizing maps (SOMs), in combination with hierarchical clustering, to identify customer groups and describe the characteristics of these clusters. The analysis was able to identify clusters of potential customers with similar preferences and gained insights into their investment patterns related to their investment portfolio and investment behavior, including their savings profile, attitude to risk-taking, and preferences for investment advice. These findings were supported by additional insights through the application of multiple correspondence analysis (MCA) describing patterns of financial instruments and portfolios. The main contribution of the research is the combined application of the machine learning methods Kohonen SOM, hierarchical clustering, and MCA for investment pattern analysis in the retail banking business.

Download Full-text

Smart Pooling: AI-powered COVID-19 testing

10.1101/2020.07.13.20152983 ◽

2020 ◽

Author(s):

María Escobar ◽

Guillaume Jeanneret ◽

Laura Bravo-Sánchez ◽

Angela Castillo ◽

Catalina Gómez ◽

...

Keyword(s):

Machine Learning ◽

World War Ii ◽

Optimal Strategies ◽

Respiratory Illness ◽

Molecular Testing ◽

Efficiency Gains ◽

Two Stage ◽

Learning Methods ◽

Pooled Testing ◽

Machine Learning Methods

Background: COVID-19 is an acute respiratory illness caused by the novel coronavirusSARS-CoV-2. The disease has rapidly spread to most countries and territories and hascaused 14.2 million confirmed infections and 602,037 deaths as of July 19th 2020. Massive molecular testing for COVID-19 has been pointed as fundamental to moderate the spread of the disease. Pooling methods can enhance the efficiency of testing, but they are viable only at very low incidences of the disease. We propose Smart Pooling, a machine learning method that uses clinical and sociodemographic data from patients to increase the efficiency of pooled molecular testing for COVID-19 by arranging samples into all-negative pools. Methods: We developed machine learning methods that estimate the probability that a sample will test positive for SARS-Cov-2 based on complementary information from the sample. We use these predictions to exclude samples predicted as positive from pools. We trained our machine learning methods on a dataset of 2000 patients tested for SARS-Cov-2 from April to July in Bogota, Colombia. Findings: Our method, Smart Pooling, shows efficiency of 306% at a disease prevalence of 5% and efficiency of 107% at disease a prevalence of up to 50%, a regime in which two-stage pooling offers marginal efficiency gains compared to individual testing. Additionally, we calculate the possible efficiency gains of one- and two-dimensional two-stage pooling strategies, and present the optimal strategies for disease prevalences up to 25%. We discuss practical limitations to conduct pooling in the laboratory. Interpretation: Pooled testing has been a theoretically alluring option to increase the coverage of diagnostics since its proposition by Dorfmann during World War II. Although there are examples of successfully using pooled testing to reduce the cost of diagnostics, its applicability has remained limited because efficiency drops rapidly as prevalence increases. Not only does our method provide a cost-effective solution to increase the coverage of testing amid the COVID-19 pandemic, but it also demonstrates that artificial intelligence can be used complementary with well-established techniques in the medical praxis.

Download Full-text

Comparison of machine learning methods for runoff forecasting in mountainous watersheds with limited data / Porównanie metod uczenia maszynowego do prognozowania spływu w zlewniach górskich na podstawie ograniczonych danych

Journal of Water and Land Development ◽

10.2478/v10025-012-0038-4 ◽

2012 ◽

Vol 17 (1) ◽

pp. 89-97 ◽

Cited By ~ 10

Author(s):

Jan Adamowski ◽

Shiv O. Prasher

Keyword(s):

Machine Learning ◽

Support Vector ◽

Limited Data ◽

Learning Methods ◽

Mountainous Regions ◽

Wavelet Networks ◽

Machine Learning Methods ◽

Antecedent Precipitation ◽

Runoff Forecasting ◽

Mountainous Watersheds

Abstract Runoff forecasting in mountainous regions with processed based models is often difficult and inaccurate due to the complexity of the rainfall-runoff relationships and difficulties involved in obtaining the required data. Machine learning models offer an alternative for runoff forecasting in these regions. This paper explores and compares two machine learning methods, support vector regression (SVR) and wavelet networks (WN) for daily runoff forecasting in the mountainous Sianji watershed located in the Himalayan region of India. The models were based on runoff, antecedent precipitation index, rainfall, and day of the year data collected over the three year period from July 1, 2001 and June 30, 2004. It was found that both the methods provided accurate results, with the best WN model slightly outperforming the best SVR model in accuracy. Both the WN and SVR methods should be tested in other mountainous watershed with limited data to further assess their suitability in forecasting.

Download Full-text