scholarly journals Estimating Time of Driver Arrival with Gradient Boosting Algorithms and Deep Neural Networks

Author(s):  
Henrik Sergoyan

Customer experience and resource management determine the degree to which transportation service providers can compete in today’s heavily saturated markets. The paper investigates and suggests a new methodology to optimize calculations for Estimated Time of Arrival (from now on ETA, meaning the time it will take for the driver to reach the designated location) based on the data provided by GG collected from rides made in 2018. GG is a transportation service providing company, and it currently uses The Open Source Routing Machine (OSRM) which exhibits significant errors in the prediction phase. This paper shows that implementing algorithms such as XGBoost, CatBoost, and Neural Networks for the said task will improve the accuracy of estimation. Paper discusses the benefits and drawbacks of each model and then considers the performance of the stacking algorithm that combines several models into one. Thus, using those techniques, final results showed that Mean Squared Error (MSE) was decreased by 54% compared to the current GG model.

Author(s):  
Moritz Feigl ◽  
Katharina Lebiedzinski ◽  
Mathew Herrnegger ◽  
Karsten Schulz

ZusammenfassungDie Fließgewässertemperatur ist ein essenzieller Umweltfaktor, der das Potenzial hat, sowohl ökologische als auch sozio-ökonomische Rahmenbedingungen im Umfeld eines Gewässers zu verändern. Um Fließgewässertemperaturen als Grundlage für effektive Anpassungsstrategien für zukünftige Veränderungen (z. B. durch den Klimawandel) berechnen zu können, sind adäquate Modellierungskonzepte notwendig. Die vorliegende Studie untersucht hierfür 6 Machine Learning-Modelle: Schrittweise Lineare Regression, Random Forest, eXtreme Gradient Boosting, Feedforward Neural Networks und zwei Arten von Recurrent Neural Networks. Die Modelle wurden an 10 österreichischen Einzugsgebieten mit unterschiedlichen physiographischen Eigenschaften und Eingangsdatenkombinationen getestet. Die Hyperparameter der angewandten Modelle wurden mittels Bayes’scher Hyperparameteroptimierung optimiert. Um die Ergebnisse mit anderen Studien vergleichbar zu machen, wurden die Vorhersagen der 6 Machine Learning-Modelle den Ergebnissen der linearen Regression und dem häufig verwendeten und bekannten Wassertemperaturmodell air2stream gegenübergestellt.Von den 6 getesteten Modellen zeigten die Feedforward Neural Networks und das eXtreme Gradient Boosting die besten Vorhersagen in jeweils 4 von 10 Einzugsgebieten. Mit einem durchschnittlichen RMSE (Wurzel der mittleren Fehlerquadratsumme; root mean squared error) von 0,55 °C konnten die getesteten Modelle die Fließgewässertemperaturen deutlich besser prognostizieren als die lineare Regression (1,55 °C) und air2stream (0,98 °C). Generell zeigten die Ergebnisse der 6 Modelle eine sehr vergleichbare Leistung mit lediglich einer mittleren Abweichung um den Medianwert von 0,08 °C zwischen den einzelnen Modellen. Im größten untersuchten Einzugsgebiet – Donau bei Kienstock – wiesen Recurrent Neural Networks die höchste Modellgüte auf, was darauf hinweist, dass sie sich am besten eignen, wenn im Einzugsgebiet Prozesse mit langfristigen Abhängigkeiten ausschlaggebend sind. Die Wahl der Hyperparameter beeinflusste die Vorhersagefähigkeit der Modelle stark, was die Bedeutung der Hyperparameteroptimierung besonders hervorhebt.Die Ergebnisse dieser Studie fassen die Bedeutung unterschiedlicher Eingangsdaten, Modelle und Trainingscharakteristiken für die Modellierung von mittleren täglichen Fließgewässertemperaturen zusammen. Gleichzeitig dient diese Studie als Basis für die Entwicklung zukünftiger Modelle für eine regionale Fließgewässertemperaturvorhersage. Die getesteten Modelle stehen im open source R‑Paket wateRtemp allen AnwenderInnen der Forschungsgemeinschaft und der Praxis zur Verfügung.


Author(s):  
Vishal Babu Siramshetty ◽  
Dac-Trung Nguyen ◽  
Natalia J. Martinez ◽  
Anton Simeonov ◽  
Noel T. Southall ◽  
...  

The rise of novel artificial intelligence methods necessitates a comparison of this wave of new approaches with classical machine learning for a typical drug discovery project. Inhibition of the potassium ion channel, whose alpha subunit is encoded by human Ether-à-go-go-Related Gene (hERG), leads to prolonged QT interval of the cardiac action potential and is a significant safety pharmacology target for the development of new medicines. Several computational approaches have been employed to develop prediction models for assessment of hERG liabilities of small molecules including recent work using deep learning methods. Here we perform a comprehensive comparison of prediction models based on classical (random forests and gradient boosting) and modern (deep neural networks and recurrent neural networks) artificial intelligence methods. The training set (~9000 compounds) was compiled by integrating hERG bioactivity data from ChEMBL database with experimental data generated from an in-house, high-throughput thallium flux assay. We utilized different molecular descriptors including the latent descriptors, which are real-valued continuous vectors derived from chemical autoencoders trained on a large chemical space (> 1.5 million compounds). The models were prospectively validated on ~840 in-house compounds screened in the same thallium flux assay. The deep neural networks performed significantly better than the classical methods with the latent descriptors. The recurrent neural networks that operate on SMILES provided highest model sensitivity. The best models were merged into a consensus model that offered superior performance compared to reference models from academic and commercial domains. Further, we shed light on the potential of artificial intelligence methods to exploit the chemistry big data and generate novel chemical representations useful in predictive modeling and tailoring new chemical space.<br>


2019 ◽  
Vol 962 ◽  
pp. 41-48
Author(s):  
Tzong Daw Wu ◽  
Jiun Shen Chen ◽  
Ching Pei Tseng ◽  
Cheng Chang Hsieh

This study presents a real-time method for determining the thickness of each layer in multilayer thin films. Artificial neural networks (ANNs) were introduced to estimate thicknesses from a transmittance spectrum. After training via theoretical spectra which were generated by thin-film optics and modified by noise, ANNs were applied to estimate the thicknesses of four-layer nanoscale films which were TiO2, Ag, Ti, and TiO2 thin films assembled sequentially on polyethylene terephthalate (PET) substrates. The results reveal that the mean squared error of the estimation is 2.6 nm2, and is accurate enough to monitor film growth in real time.


Proceedings ◽  
2020 ◽  
Vol 59 (1) ◽  
pp. 2
Author(s):  
Benoit Figuet ◽  
Raphael Monstein ◽  
Michael Felux

In this paper, we present an aircraft localization solution developed in the context of the Aircraft Localization Competition and applied to the OpenSky Network real-world ADS-B data. The developed solution is based on a combination of machine learning and multilateration using data provided by time synchronized ground receivers. A gradient boosting regression technique is used to obtain an estimate of the geometric altitude of the aircraft, as well as a first guess of the 2D aircraft position. Then, a triplet-wise and an all-in-view multilateration technique are implemented to obtain an accurate estimate of the aircraft latitude and longitude. A sensitivity analysis of the accuracy as a function of the number of receivers is conducted and used to optimize the proposed solution. The obtained predictions have an accuracy below 25 m for the 2D root mean squared error and below 35 m for the geometric altitude.


2017 ◽  
Vol 3 (1) ◽  
pp. 10
Author(s):  
Debby E. Sondakh

Classification has been considered as an important tool utilized for the extraction of useful information from healthcare dataset. It may be applied for recognition of disease over symptoms. This paper aims to compare and evaluate different approaches of neural networks classification algorithms for healthcare datasets. The algorithms considered here are Multilayer Perceptron, Radial Basis Function, and Voted Perceptron which are tested based on resulted classifiers accuracy, precision, mean absolute error and root mean squared error rates, and classifier training time. All the algorithms are applied for five multivariate healthcare datasets, Echocardiogram, SPECT Heart, Chronic Kidney Disease, Mammographic Mass, and EEG Eye State datasets. Among the three algorithms, this study concludes the best algorithm for the chosen datasets is Multilayer Perceptron. It achieves the highest for all performance parameters tested. It can produce high accuracy classifier model with low error rate, but suffer in training time especially of large dataset. Voted Perceptron performance is the lowest in all parameters tested. For further research, an investigation may be conducted to analyze whether the number of hidden layer in Multilayer Perceptron’s architecture has a significant impact on the training time.


2021 ◽  
Author(s):  
Mengbo Guo ◽  
Xuyang Xu ◽  
Han Xie

Density functional theory (DFT) is a ubiquitous first-principles method, but the approximate nature of the exchange-correlation functional poses an inherent limitation for the accuracy of various computed properties. In this context, surrogate models based on machine learning have the potential to provide a more efficient and physically meaningful understanding of electronic properties, such as the band gap. Here, we construct a gradient boosting regression (GBR) model for prediction of the band gap of binary compounds from simple physical descriptors, using a dataset of over 4000 DFT-computed band gaps. Out of 27 features, electronegativity, periodic group, and highest occupied energy level exhibit the highest importance score, consistent with the underlying physics of the electronic structure. We obtain a model accuracy of 0.81 and root mean squared error of 0.26 eV using the top five features, achieving accuracy comparable to previously reported values but employing less number of features. Our work presents a rapid and interpretable prediction model for solid-state band gap with high fidelity to DFT and can be extended beyond binary materials considered in this study.


Author(s):  
Leonardo Fabio León Marenco ◽  
Luiza Pereira Oliveira ◽  
Daniella Lopez Vale ◽  
Maiara Oliveira Salles

Abstract An artificial neural network was used to build models caple of predicting and quantifying vodka adulteration with methanol and/or tap water. A voltammetric electronic tongue based on gold and copper microelectrodes was used, and 310 analyses were performed. Vodkas were adulterated with tap water (5 to 50% (v/v)), methanol (1 to 13% (v/v)), and with a fixed addition of 5% methanol and tap water varying from 5 to 50% (v/v). The classification model showed 99.5% precision, and it correctly predicted the type of adulterant in all samples. Regarding the regression model, the root mean squared error was 3.464% and 0.535% for the water and methanol addition, respectively, and the prediction of the adulterant content presented an R2 0.9511 for methanol and 0.9831 for water adulteration.


Volume 1 ◽  
2004 ◽  
Author(s):  
I. Z. Mat Darus ◽  
M. O. Tokhi ◽  
S. Z. Mohd. Hashim

This paper investigates the utilisation of feedforward and recurrent neural networks for dynamic modelling of a flexible plate structure. Neuro-modelling techniques are used for non-parametric identification of the flexible plate structure based on one-step-ahead prediction. A multi layer perceptron (MLP) and Elman neural networks are designed to characterise the dynamic behaviour of the flexible plate. Results of the modelling techniques are validated through a range of tests including input/output mapping, training and test validation, mean-squared error and correlation tests. Results are presented in both time and frequency domains. Comparative performance assessments of both neuro-modelling approaches in terms of mean-squared error and estimation of the resonance modes of the system are carried out. It is noted that both techniques have been able to detect the first five vibration modes of the system successfully. Investigations also signify the advantage of a recurrent Elman network over an MLP feedforward network in modelling the flexible plate structure.


Genes ◽  
2019 ◽  
Vol 10 (11) ◽  
pp. 862
Author(s):  
Tong Liu ◽  
Zheng Wang

We present a deep-learning package named HiCNN2 to learn the mapping between low-resolution and high-resolution Hi-C (a technique for capturing genome-wide chromatin interactions) data, which can enhance the resolution of Hi-C interaction matrices. The HiCNN2 package includes three methods each with a different deep learning architecture: HiCNN2-1 is based on one single convolutional neural network (ConvNet); HiCNN2-2 consists of an ensemble of two different ConvNets; and HiCNN2-3 is an ensemble of three different ConvNets. Our evaluation results indicate that HiCNN2-enhanced high-resolution Hi-C data achieve smaller mean squared error and higher Pearson’s correlation coefficients with experimental high-resolution Hi-C data compared with existing methods HiCPlus and HiCNN. Moreover, all of the three HiCNN2 methods can recover more significant interactions detected by Fit-Hi-C compared to HiCPlus and HiCNN. Based on our evaluation results, we would recommend using HiCNN2-1 and HiCNN2-3 if recovering more significant interactions from Hi-C data is of interest, and HiCNN2-2 and HiCNN if the goal is to achieve higher reproducibility scores between the enhanced Hi-C matrix and the real high-resolution Hi-C matrix.


Author(s):  
Gebreab K. Zewdie ◽  
David J. Lary ◽  
Estelle Levetin ◽  
Gemechu F. Garuma

Allergies to airborne pollen are a significant issue affecting millions of Americans. Consequently, accurately predicting the daily concentration of airborne pollen is of significant public benefit in providing timely alerts. This study presents a method for the robust estimation of the concentration of airborne Ambrosia pollen using a suite of machine learning approaches including deep learning and ensemble learners. Each of these machine learning approaches utilize data from the European Centre for Medium-Range Weather Forecasts (ECMWF) atmospheric weather and land surface reanalysis. The machine learning approaches used for developing a suite of empirical models are deep neural networks, extreme gradient boosting, random forests and Bayesian ridge regression methods for developing our predictive model. The training data included twenty-four years of daily pollen concentration measurements together with ECMWF weather and land surface reanalysis data from 1987 to 2011 is used to develop the machine learning predictive models. The last six years of the dataset from 2012 to 2017 is used to independently test the performance of the machine learning models. The correlation coefficients between the estimated and actual pollen abundance for the independent validation datasets for the deep neural networks, random forest, extreme gradient boosting and Bayesian ridge were 0.82, 0.81, 0.81 and 0.75 respectively, showing that machine learning can be used to effectively forecast the concentrations of airborne pollen.


Sign in / Sign up

Export Citation Format

Share Document