Application Of Machine Learning For Prediction Research Articles

Machine learning application for prediction of surgical site infection after posterior cervical surgery.

Surgical site infection (SSI) is one of the most common complications of posterior cervical surgery. It is difficult to diagnose in the early stage and may lead to severe consequences such as wound dehiscence and central nervous system infection. This retrospective study included patients who underwent posterior cervical surgery at The Second Affiliated Hospital and Yuying Childrens Hospital of Wenzhou Medical University from September 2018 to June 2022. We employed several machine learning methods, such as the gradient boosting (GB), random forests (RF), artificial neural network (ANN) and other popular machine learning models. To minimize the variability introduced by random splitting, the results underwent 10-fold cross-validation repeated 10 times. Five measurements were averaged across 10 repetitions with 10-fold cross-validation, the RF model achieved the highest AUROC (0.9916), specificity (0.9890) and precision (0.9759). The GB model achieved the highest sensitivity (0.9535) and the KNN achieved the highest sensitivity (0.9958). The application of machine learning techniques facilitated the development of a precise model for predicting SSI after posterior cervical surgery. This dynamic model can be served as a valuable tool for clinicians and patients to assess SSI risk and prevent it in clinical practice.

Machine learning application for prediction of sonic wave transit time - A case of Niger Delta basin

Sonic wave transit times are the major logs for estimating pertinent geomechanical parameters including overburden stress, pore pressure, effective stress, and unconfined compressible strength among others. These logs contain much important information about subsurface formation. However, in most oil and gas exploration wells, these logs are not usually acquired in all the wells which can be attributed to the high cost involved in logging. The application of the data-driven ML model has rarely been used for compressional and shear sonic logs (DTP and DTS) prediction within the Niger Delta basin where many reports of geomechanical instabilities have been reported. In this study, seven machine learning algorithms namely linear regression (LR), K-nearest neighbour (KNN), support vector regressor (SVR), random forest regressor (RFR), Extreme Gradient Boosting (XGBoost), Categorical Boosting (CatBoost), and back propagation artificial neural network (BPANN) were employed for prediction of DTP and DTS from conventional wireline logs data and their results were compared. For DTP prediction, the best model achieved a coefficient of determination (R2) value of 90 % using CatBoost. Also, CatBoost produced the model for DTS prediction with an R2 value of 95 %. A comparison results of the seven ML algorithms showed that six models performed really well for DTP and DTS predictions with a value of R2 > 80 % and 90 % respectively. In contrast, the worst performance models are LR for DTP prediction and SVR for DTS prediction. The blind testing carried out on some selected wells affirmed the efficiency of the proposed technique producing R2 values ranging from 92 to 94 % for DTP prediction and 94–99 % for DTS prediction. This present study confirmed the ability of building efficient ML models for estimating and predicting DTP and DTS which can be used to estimate geomechanical parameters in wells where both are not available.

Machine learning application for prediction of sapphire crystals defects

We investigate the impact of different numbers of positive and negative examples on machine learning for sapphire crystals defects prediction. We obtain the models of crystal growth parameters influence on the sapphire crystal growth. For example, these models allow predicting the defects that occur due to local overcooling of crucible walls in the thermal node leading to the accelerated crystal growth. We also develop the prediction models for obtained crystal weight, blocks, cracks, bubbles formation, and total defect characteristics. The models were trained on all data sets and later tested for generalization on testing sets, which did not overlap the training set. During training and testing, we find the recall, precision of prediction and analyze the correlation among the features. The results have shown that the precision of neural network method for predicting defects formed by local overcooling of the crucible reached 0.94.

Machine learning application for prediction of locoregional recurrences in early oral tongue cancer: a Web-based prognostic tool

Estimation of risk of recurrence in early-stage oral tongue squamous cell carcinoma (OTSCC) remains a challenge in the field of head and neck oncology. We examined the use of artificial neural networks (ANNs) to predict recurrences in early-stage OTSCC. A Web-based tool available for public use was also developed. A feedforward neural network was trained for prediction of locoregional recurrences in early OTSCC. The trained network was used to evaluate several prognostic parameters (age, gender, T stage, WHO histologic grade, depth of invasion, tumor budding, worst pattern of invasion, perineural invasion, and lymphocytic host response). Our neural network model identified tumor budding and depth of invasion as the most important prognosticators to predict locoregional recurrence. The accuracy of the neural network was 92.7%, which was higher than that of the logistic regression model (86.5%). Our online tool provided 88.2% accuracy, 71.2% sensitivity, and 98.9% specificity. In conclusion, ANN seems to offer a unique decision-making support predicting recurrences and thus adding value for the management of early OTSCC. To the best of our knowledge, this is the first study that applied ANN for prediction of recurrence in early OTSCC and provided a Web-based tool.

Novel machine learning application for prediction of membrane insertion potential of cell-penetrating peptides.

Cell-penetrating peptides (CPPs) are often used as transporter systems to deliver various therapeutic agents into the cell. We developed a novel machine learning application which can quantitatively screen the insertion/interaction potential of various CPPs into three model phospholipid monolayers. An artificial neural network (ANN) was designed, trained, and ultimately tested on an external dataset using Langmuir experimental data for 13 CPPs (hydrophilic and amphiphilic) together with various features related to the insertion/interaction efficiency of CPPs. The trained ANN provided accurate predictions of the maximum change in surface pressure of CPPs when injected below three membrane models at pH 7.4. The accuracy of predictions was high for the dataset which was used to construct the model (r2 = 0.986) as well as for the external "prospective" dataset (r2 = 0.969). In conclusion, this study demonstrates the promising potential of ANNs for screening the insertion potential of CPPs into membrane models for efficient intracellular delivery of therapeutic agents.