No abstract

Performances of smartphones are profoundly affected by battery life. Maximizing the amount of usage of energy is essential to extend battery life. However, developers might concentrate more on the functionality of applications while ignoring the energy bugs that drain the battery during the development process. There are no quantitative approaches to detect these energy bugs introduced in this fast-paced development process. In this paper, we employ a system-call-based approach to develop a power consumption model for Android devices. Data that measure the energy consumption of mobile devices under different testing scenarios with the number of triggered system calls are utilized in the model training process. A balanced recursive feature elimination with cross-validation approach is proposed to select and rank the importance of the different system calls. Seven machine learning models are trained over the selected features with cross-validation and hyper-parameter tuning technique, where linear regression with the Lasso regularization outperforms all the other models. Then, the model is evaluated on the data set that measures the energy consumption on different revision history of the selected apps. The results show that the optimized Lasso model could detect energy bugs in the revision history of various applications. Optimization strategies are provided based on the selected features.

Lithology identification is vital for reservoir exploration and petroleum engineering. Recently, there has been growing interest in using an intelligent logging approach for lithology classification. Machine learning has emerged as a powerful tool in inferring lithology types with the logging curves. However, well logs are susceptible to logging parameter manual entry, borehole conditions and tool calibrations. Most studies in the field of lithology classification with machine learning approaches have focused only on improving the prediction accuracy of classifiers. Also, a model trained in one location is not reusable in a new location due to different data distributions. In this paper, a unified framework is provided for training a multi-class lithology classification model for a data set with outlier data. In this paper, a coarse-to-fine framework that combines outlier detection, multi-class classification with an extremely randomized tree-based classifier is proposed to solve these issues. An unsupervised learning approach is used to detect the outliers in the data set. Then a coarse-to-fine inference procedure is used to infer the lithology class with an extremely randomized tree classifier. Two real-world data sets of well-logging are used to demonstrate the effectiveness of the proposed framework. Comparisons are conducted with some baseline machine learning classifiers, namely random forest, gradient tree boosting, and xgboosting. Results show that the proposed framework has higher prediction accuracy in sandstones compared with other approaches.

Recently, there has been growing interest in improving the efficiency and accuracy of the Indoor Positioning System (IPS). The Received Signal Strength- (RSS-) based fingerprinting technique is essential for indoor localization. However, it is challenging to estimate the indoor position based on RSS’s measurement under the complex indoor environment. This paper evaluates three machine learning approaches and Gaussian Process (GP) regression with three different kernels to get the best indoor positioning model. The hyperparameter tuning technique is used to select the optimum parameter set for each model. Experiments are carried out with RSS data from seven access points (AP). Results show that GP with a rational quadratic kernel and eXtreme gradient tree boosting model has the best positioning accuracy compared to other models. In contrast, the eXtreme gradient tree boosting model could achieve higher positioning accuracy with smaller training size and fewer access points.

Lithology identification is an indispensable part in geological research and petroleum engineering study. In recent years, several mathematical approaches have been used to improve the accuracy of lithology classification. Based on our earlier work that assessed machine learning models on formation lithology classification, we optimize the boosting approaches to improve the classification ability of our boosting models with the data collected from the Daniudi gas field and Hangjinqi gas field. Three boosting models, namely, AdaBoost, Gradient Tree Boosting, and eXtreme Gradient Boosting, are evaluated with 5-fold cross validation. Regularization is applied to the Gradient Tree Boosting and eXtreme Gradient Boosting to avoid overfitting. After adapting the hyperparameter tuning approach on each boosting model to optimize the parameter set, we use stacking to combine the three optimized models to improve the classification accuracy. Results suggest that the optimized stacked boosting model has better performance concerning the evaluation matrix such as precision, recall, and f1 score compared with the single optimized boosting model. Confusion matrix also shows that the stacked model has better performance in distinguishing sandstone classes.

Lithology identification is critical in the interpretation of well-logging data for petroleum exploration and development. However, the limited availability of labeled well-logging data for machine learning model training can lead to compromised accuracy in lithology classification models. Here, we propose a semi-supervised lithology identification model to overcome this challenge. Our framework consists of Bayesian optimization for tuning ensemble algorithms, including random forest, gradient boosting decision tree, extremely randomized trees, and adaptive boosting, to establish a high-quality baseline model for semi-supervised learning. We also employ a self-training strategy to increase the number of labeled samples in the training set and use the predicted label with the highest confidence as a pseudo-label to reduce the accumulation of deviation caused by incorrect pseudo-labels. Our semi-supervised coarse-to-fine framework improves rock classification accuracy, particularly for sandstone. Testing our model on well-logging data from two real regions, we found that the ExtraRF-based semi-supervised model in the HGF area performs the best, with a maximum classification accuracy of 91.6$$\%$$ % , which is 5$$\%$$ % higher than the original coarse-to-fine model without using Bayesian optimization and pseudo-labeling techniques.