Combination Of Individual Models Research Articles

S3: An AI-Enabled User Continuous Authentication for Smartphones Based on Sensors, Statistics and Speaker Information

Continuous authentication systems have been proposed as a promising solution to authenticate users in smartphones in a non-intrusive way. However, current systems have important weaknesses related to the amount of data or time needed to build precise user profiles, together with high rates of false alerts. Voice is a powerful dimension for identifying subjects but its suitability and importance have not been deeply analyzed regarding its inclusion in continuous authentication systems. This work presents the S3 platform, an artificial intelligence-enabled continuous authentication system that combines data from sensors, applications statistics and voice to authenticate users in smartphones. Experiments have tested the relevance of each kind of data, explored different strategies to combine them, and determined how many days of training are needed to obtain good enough profiles. Results showed that voice is much more relevant than sensors and applications statistics when building a precise authenticating system, and the combination of individual models was the best strategy. Finally, the S3 platform reached a good performance with only five days of use available for training the users’ profiles. As an additional contribution, a dataset with 21 volunteers interacting freely with their smartphones for more than sixty days has been created and made available to the community.

Optimal group selection algorithm in air quality index forecasting via cooperative information criterion

With the increasing concern of government and the public on air quality, accurately forecasting air quality index is important in guiding pollution control and protecting public health at an early stage. The main purpose of this research is to develop optimal combined forecast for predicting air pollutant. Therefore, a selection method based on cooperative information criterion is proposed to determine the optimal forecasting group from multiple individual forecasting models. The method excludes an individual forecast from a group if the combination produced by the forecast and the subgroup containing other forecasts is not a superior combination. The daily air quality data collected from Hefei are utilized as a case to verify the effectiveness of proposed selection approach. In the first stage of empirical research, four major air pollutants which influence Hefei’s air quality significantly are identified. Then different optimal forecasting groups for four major pollutants are selected from ten available models respectively. The experimental results demonstrate that the combination of individual models from the selected optimal group outperforms both the best individual model and the combination of all available models. Finally, we can obtain the forecasting values of overall air quality index by integrating the information provided by four major air pollutants.

Parallel hybridization of series (PHOS) models for time series forecasting

Literature indicates that many efforts have been conducted toward the development of forecasting models with a high degree of accuracy. Combining different models is known as a powerful alternative to access more reliable and more accurate results than single models. Given the great importance of hybridization theory, various hybrid models have been proposed in the literature of time series forecasting. Series hybrid approaches are one of the most well study and the most widely used hybridization models, in which components are sequentially applied. However, according to the modeling essence of this hybrid structure, the performance of the series hybrid models is directly dependent on the order of using components and modeling sequence. Besides, selecting the best modeling order that can yield the best performance in all situations is a problematic theoretical as well as practical task. Thus, the main purpose of this paper is to eliminate the drawback of series models, regarding modeling order selection using a parallel hybridization schema, which is addressed for the first time. The core principle of the proposed parallel hybridization of series (PHOS) models is to improve the series hybrid model’s forecasting accuracy by integrating two hybrid structures in contrast to existing hybrid models, which emphasize only the combination of individual models. The proposed model decomposes the original time series into two linear and nonlinear parts and uses the autoregressive integrated moving average (ARIMA) and multilayer perceptron neural network (MLP) models to model underlying patterns, incorporating two series models including ARIMA–MLP and MLP–ARIMA. Finally, the series models are integrated as components of the parallel hybridization scheme. Moreover, an ordinary least square algorithm is developed to determine the optimal weights of these two components. Three benchmark data sets, including the closing of the DAX index, the closing of the Nikkei 225 index (N225), and the opening of the Dow Jones Industrial Average Index, are used for empirical analysis and verifying the effectiveness of the PHOS model. The empirical analysis indicates that the PHOS model can improve the forecasting performance of both series ARIMA–MLP and MLP–ARIMA models as well as individual models and some parallel hybrid models.

Forecasting the real price of oil - Time-variation and forecast combination

This paper sheds light on the questions whether it is possible to generate an accurate forecast of the real price of oil and how it can be improved using forecast combinations. For this reason, the following paper will investigate the out-of-sample performance of seven individual forecasting models. The results show that it is possible to construct better forecasts compared to a no-change benchmark for horizons up to 24 months with gains in the MSPE ratio as high as 25%. In addition, some of the existing models will be extended, e.g. the U.S. inventories model by introducing more suitable real-time measures for the Brent crude oil price and the VAR model of the global oil market by using different measures for the economic activity. Furthermore, the time performance investigated by constructing recursively estimated MSPE ratios discovers potential weaknesses of the used models. Hence, several different combination approaches are tested with the goal of demonstrating that a combination of individual models is beneficial for the forecasting performance. Thereby, a combination consisting of four models has proven to have a lower MSPE ratio than the best individual models over the medium run and, in addition, to be remarkably stable over time.

Automatic selection for general surrogate models

In design engineering problems, the use of surrogate models (also called metamodels) instead of expensive simulations have become very popular. Surrogate models include individual models (regression, kriging, neural network...) or a combination of individual models often called aggregation or ensemble. Since different surrogate types with various tunings are available, users often struggle to choose the most suitable one for a given problem. Thus, there is a great interest in automatic selection algorithms. In this paper, we introduce a universal criterion that can be applied to any type of surrogate models. It is composed of three complementary components measuring the quality of general surrogate models: internal accuracy (on design points), predictive performance (cross-validation) and a roughness penalty. Based on this criterion, we propose two automatic selection algorithms. The first selection scheme finds the optimal ensemble of a set of given surrogate models. The second selection scheme further explores the space of surrogate models by using an evolutionary algorithm where each individual is a surrogate model. Finally, the performances of the algorithms are illustrated on 15 classical test functions and compared to different individual surrogate models. The results show the efficiency of our approach. In particular, we observe that the three components of the proposed criterion act all together to improve accuracy and limit over-fitting.

Forecasting the Real Price of Oil - Time-Variation and Forecast Combination

This paper sheds light on the questions whether it is possible to generate an accurate forecast of the real price of oil and how it can be improved using forecast combinations. For this reason, my work will investigate the out-of-sample performance of thirteen individual forecasting models. The results show that it is possible to construct better forecasts compared to a no-change benchmark for horizons up to 24 months with gains in the MSPE ratio as high as 25%. In addition, I will extend some of the existing models, e.g the U.S. inventories model by introducing more suitable real time measures for the Brent crude oil price and the VAR model of the global oil market by using different measures for the economic activity. Furthermore, the time performance investigated by constructing recursively estimated MSPE ratios discovers potential weaknesses of the used models. Hence, several different combination approaches are tested with the goal of demonstrating that a combination of individual models is beneficial for the forecasting performance. Thereby, a combination consisting of four models has proven to have a lower MSPE ratio than the best individual models over the medium run and, in addition, to be remarkably stable over time.

Intelligent Models to Identification and Treatment of Outliers in Electrical Load Data

The Electrical Load data are stored in each time interval generating big databases with high dimensional data. Each data stored contains significant information that can assist the planning and operation of electrical system. In the data analysis step, many aspects must be considered such as the data consistency and the identification and treatment of outliers. This is a critical step because data quality is directly reflected in the results of the planning and operation of electrical system. This paper proposes two models for the identification and treatment of outliers in electrical load data. The first model was built using the ensemble technique through a combination of individual models. The second model was created from an expert system that uses a rules database to detect outliers. The processing of the outliers detected is conducted through a combination of non-outliers load in the same time interval. To evaluate the performance, the models were applied in a historical load database measured in the Northeast of Brazil during year of 2006. The results showed that the proposed models showed satisfactory results in terms of detection as well in the treatment of outliers.

A combination selection algorithm on forecasting

It is widely accepted in forecasting that a combination model can improve forecasting accuracy. One important challenge is how to select the optimal subset of individual models from all available models without having to try all possible combinations of these models. This paper proposes an optimal subset selection algorithm from all individual models using information theory. The experimental results in tourism demand forecasting demonstrate that the combination of the individual models from the selected optimal subset significantly outperforms the combination of all available individual models. The proposed optimal subset selection algorithm provides a theoretical approach rather than experimental assessments which dominate literature.

Development of a highly specific ensemble of topological models for early identification of P‐glycoprotein substrates

P‐glycoprotein (Pgp) is an ATP‐dependent efflux transporter protein associated with multidrug resistance in several diseases such as cancer, epilepsy and AIDS. It is preferentially expressed in organs and tissues that function as a barrier (e.g. the gut walls or the blood–brain barrier) or promote the elimination of xenobiotics from the organism (e.g. liver and kidney). Pgp limits drug bioavailability; thus, the recognition of Pgp substrates at the early stages of the drug development cycle is essential for the development of new chemotherapeutic agents to deal with multidrug resistance issues. Here we present the development of several classifier models based on topological descriptors to identify potential Pgp substrates, aimed to be applied as secondary filter in virtual screening campaigns. Receiver Operating characteristic (ROC) curves show that combination of individual models, through data fusion, in a three‐model ensemble, allows attaining higher areas under the curve and an overall better behavior in terms of sensitivity or specificity. The individual discriminant functions (dfs) presented have a performance similar to that of the previously reported models and, remarkably, our models only include low‐dimensional (up to 2D) molecular descriptors, which makes them adequate for the virtual screening of increasingly large virtual chemical repositories. Copyright © 2011 John Wiley & Sons, Ltd.