The autofeat Python Library for Automated Feature Engineering and Selection

Horn, Franziska; Pack, Robert C.; Rieger, Michael L.

doi:10.1007/978-3-030-43823-4_10

Cited by 49 publications

(30 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The R 2 for linear regression, however, increased considerably from 0.80 to 0.96. This is rather expected considering that even on benchmark datasets, the model showed remarkable improvements on linear regression tasks but could not outperform the state-ofthe-art RF regression models 36 .…”

Section: Methodsmentioning

confidence: 91%

“…In fact, an earlier study on protein-ligand binding affinity also exhibited that predictions based on RF and decision trees consistently outperformed linear regression models 35 . In order to further augment the predictive power of our models, we implemented an automated feature engineering and selection using the autofeat library of python 36 . It is a framework inspired by the SISSO algorithm 37 that automatically generates a large number of non-linear features from the input descriptors and then selects the most informative of them as additional features.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Applied machine learning for predicting the lanthanide-ligand binding affinities

Chaube

Srinivasan

Rai

2020

Sci Rep

View full text Add to dashboard Cite

Binding affinities of metal-ligand complexes are central to a multitude of applications like drug design, chelation therapy, designing reagents for solvent extraction etc. While state-of-the-art molecular modelling approaches are usually employed to gather structural and chemical insights about the metal complexation with ligands, their computational cost and the limited ability to predict metalligand stability constants with reasonable accuracy, renders them impractical to screen large chemical spaces. In this context, leveraging vast amounts of experimental data to learn the metal-binding affinities of ligands becomes a promising alternative. Here, we develop a machine learning framework for predicting binding affinities (logK 1) of lanthanide cations with several structurally diverse molecular ligands. Six supervised machine learning algorithms-Random Forest (RF), k-Nearest Neighbours (KNN), Support Vector Machines (SVM), Kernel Ridge Regression (KRR), Multi Layered Perceptrons (MLP) and Adaptive Boosting (AdaBoost)-were trained on a dataset comprising thousands of experimental values of logK 1 and validated in an external 10-folds cross-validation procedure. This was followed by a thorough feature engineering and feature importance analysis to identify the molecular, metallic and solvent features most relevant to binding affinity prediction, along with an evaluation of performance metrics against the dimensionality of feature space. Having demonstrated the excellent predictive ability of our framework, we utilized the best performing AdaBoost model to predict the logK 1 values of lanthanide cations with nearly 71 million compounds present in the PubChem database. Our methodology opens up an opportunity for significantly accelerating screening and design of ligands for various targeted applications, from vast chemical spaces. Rare Earth Elements (REEs), that constitute the lanthanide block of the periodic table, together with Yttrium and Scandium, lie at the heart of many modern technologies in diverse fields ranging from health care to clean energy applications 1. With increasing adoption of clean and energy efficient technologies, the demand for REEs is expected to grow manifold in the coming years 2. Although conventional mining remains the primary source of global REE supply currently 3 , owing to the huge quantities of electronic waste (e-waste) generated, REE recovery from e-wastes becomes a promising secondary source of these critical elements 4. Much of the metal processing industry relies upon hydrometallurgical operations such as liquid-liquid extraction (LLE) to recover the target element 5. The success of an LLE operation depends critically on the choice of ligands that can selectively bind to one or more target metal ions and transport them into an oil phase in contact with an aqueous phase which originally contained the metal ions. Thus, successful recovery of REEs from e-wastes calls for the design of ligands with a high affinity for one or more target lanthanide ions. The binding strength of a ...

show abstract

Section: Methodsmentioning

confidence: 91%

Section: Methodsmentioning

confidence: 99%

Applied machine learning for predicting the lanthanide-ligand binding affinities

Chaube

Srinivasan

Rai

2020

Sci Rep

View full text Add to dashboard Cite

show abstract

“…However, in the presence of manually selected features it could be feared that important information on the gait was in advance omitted (Dindorf et al, 2020). Therefore, automated feature selection methods are an alternative approach, providing discrete as well as time series parameters (Christ et al, 2018;Horn et al, 2020). However, these methods might yield abstract features that are difficult to interpret and therefore do not allow a straight forward deduction of a treatment approach.…”

Section: Tablementioning

confidence: 99%

Automated detection and explainability of pathological gait patterns using a one-class support vector machine trained on inertial measurement unit based gait data

Teufl

Taetz

Miezal

et al. 2021

Clinical Biomechanics

View full text Add to dashboard Cite

Background: Machine learning approaches for the classification of pathological gait based on kinematic data, e.g. derived from inertial sensors, are commonly used in terms of a multi-class classification problem. However, there is a lack of research regarding one-class classifiers that are independent of certain pathologies. Therefore, it was the aim of this work to design a one-class classifier based on healthy norm-data that provides not only a prediction probability but rather an explanation of the classification decision, increasing the acceptance of this machine learning approach. Methods: The inertial sensor based gait kinematics of 25 healthy subjects was employed to train a one-class support vector machine. 25 healthy subjects, 20 patients after total hip arthroplasty and one transfemoral amputee served to validate the classifier. Prediction probabilities and feature importance scores were estimated for each subject. Findings: The support vector machine predicted 100% of the patients as outliers from the healthy group. Three healthy subjects were predicted as outliers. The feature importance calculation revealed the hip in the sagittal plane as most relevant feature concerning the group after total hip arthroplasty. For the misclassified healthy subject with the lowest probability score the knee flexion and the pelvis obliquity were identified. Interpretation: The support vector machine seems a useful tool to identify outliers from a healthy norm-group. The feature importance examination proved to provide valuable information on the musculoskeletal status of the subjects. In this combination, the present approach could be employed in various disciplines to identify abnormal gait and suggest subsequent training.

show abstract

“…," a python framework for automatic features generation, is used that automatically generates non-linear features from input and then selects important ones from generated features [35]. This library uses combination of two or more features to generate new ones.…”

Section: B Feature Engineeringmentioning

confidence: 99%

A Novel Optimized Case-Based Reasoning Approach With K-Means Clustering and Genetic Algorithm for Predicting Multi-Class Workload Characterization in Autonomic Database and Data Warehouse System

2020

View full text Add to dashboard Cite

Data management systems are essential elements for any organization which is dealing with large volume of data now a days. Due to increase in data volume, and its complexities, it has become more challenging job for workload management system to maintain its performance. So, there is a need of such a system that can autonomically deal with such complexities with less or without human involvement. Performance of these systems can be improved by making the systems well-aware about the workload entering into the system. The workload of a prevalent typical database and data warehouse system can be characterized into three types that is Online Transaction Processing (OLTP), Decision Support Systems (DSS) and Mixed type of workload. Currently, autonomic characterization of workload into a binary class such as OLTP and DSS is being carried out as reported in the literature, however, characterizing the workload into three types that refers to a multi-class classification problem is relatively a more challenging task. In this study, we propose a novel optimized Case-based Reasoning (CBR) approach based on clustering for autonomically characterizing the workload into multi-class types before entering into the system. We implement four phases of CBR along with case-base generation and map it to the elements of autonomic MAPE-K model. In Retrieve phase, k-means clustering is used for enhancing retrieval efficiency and workload types predictions are made in Reuse phase. Genetic Algorithm is used in Revise and Adapt phase of CBR. Few autonomic self_* characteristics are incorporated to make it autonomic. We performed various experiments and results show that the proposed model outperforms in prediction as compared to existing approaches. We performed posthoc test for the validation of results in comparison with other machine learning classifiers using the Friedman test that show that the proposed model stands out as the best classifier. INDEX TERMS Workload management system, workload characterization, case-based reasoning, autonomic systems, Genetic Algorithm.

show abstract

The autofeat Python Library for Automated Feature Engineering and Selection

Cited by 49 publications

References 28 publications

Applied machine learning for predicting the lanthanide-ligand binding affinities

Applied machine learning for predicting the lanthanide-ligand binding affinities

Automated detection and explainability of pathological gait patterns using a one-class support vector machine trained on inertial measurement unit based gait data

A Novel Optimized Case-Based Reasoning Approach With K-Means Clustering and Genetic Algorithm for Predicting Multi-Class Workload Characterization in Autonomic Database and Data Warehouse System

Contact Info

Product

Resources

About