In Fig. 2, the Fig. 2(d), 2(e) and 2(f) was missing from this article; Fig. 2 appeared as shown below, a Data preprocessing results across Datasets. b Top 10 feature importance score across datasets. c Comparison of hyperparameter tuning methods across algorithm (Accuracy). d Cross validation accuracy comparison for hyperparameter turing methods. e ROC curves for Multiple Datasets. f Correlation Matix (Heatmap) between top 10 features. The figure(s) should have appeared as shown below. a Data preprocessing results across Datasets. b Top 10 feature importance score across datasets. c Comparison of hyperparameter tuning methods across algorithm (Accuracy). d Cross validation accuracy comparison for hyperparameter turing methods. e ROC curves for Multiple Datasets. f Correlation Matix (Heatmap) between top 10 features. The original article has been corrected.

Automated Machine Learning (AutoML) systems are designed to overcome the complexity of machine learning (ML) through creating pipeline. These pipelines may be in constant-form or flexible depending on the requirement of a particular task. This work introduces an Optimized AutoML framework for hyperparameter tuning (OAFH) that addresses AutoML pipeline that integrates diverse ML algorithms such as SVM, KNN, Decision Tree and Neural Network with hyperparameter optimization strategies, such as GridSearchCV, Hyperopt, and Optunity. The proposed OAFH, automates all the steps of feature engineering, data preprocessing includes missing data, scaling features, and data balancing through SMOTE. Furthermore, performance is observed for best algorithm in terms of Precision, Recall, F1-score, Accuracy and MCC. Thereafter, comparative study is drawn with conventional autoML models such as EVOSA, FEDOT, AutoGluon and H2O, along with the proposed OAFH. The experimental results revealed that the proposed OAFH outperformed the conventional methods and reported an accuracy of 78% to 100% on 12 conventional datasets. This study shows how powerful AutoML systems can become in standalone usage for data analysis without the need of having precise ML knowledge.