Prediction of spontaneous combustion susceptibility of coal seams based on coal intrinsic properties using various machine learning tools.

Spontaneous combustion of coal leading to mine fire is a major problem in most of the coal mining countries in the world. It causes major loss to the Indian economy. The liability of coal to spontaneous combustion varies from place to place and mainly depends on the coal intrinsic properties and other geo-mining factors. Hence, the prediction of spontaneous combustion susceptibility of coal is of utmost importance for preventing the risk of fire in coal mines and utility sectors. Machine learning tools are pivotal in system improvements in relation to the statistical analysis of experimental results. Wet oxidation potential (WOP) of coal determined in the laboratory is one of the most relied indices used for assessing the spontaneous combustion susceptibility of coal. In this study, multiple linear regression (MLR) and five different machine learning (ML) techniques, such as Support Vector Regression (SVR), Artificial Neural Network (ANN), Random Forest (RF), Gradient Boosting (GB) and Extreme Gradient Boost (XGB) algorithms, were used to predict the spontaneous combustion susceptibility (WOP) of coal seams based on the coal intrinsic properties. The results derived from the models were compared with the experimental data. The results indicated excellent prediction accuracy and ease of interpretation of tree-based ensemble algorithms, like Random Forest, Gradient Boosting and Extreme Gradient Boosting. The MLR exhibited the least while XGB demonstrated the highest predictive performance. The developed XGB achieved R2 of 0.9879, RMSE of 4.364 and VAF of 84.28%. In addition, the results of sensitivity analysis showed that the volatile matter is most sensitive to the changes in WOP of coal samples considered in the study. Thus, during spontaneous combustion modelling and simulation, volatile matter can be used as the most influential parameter for assessing the fire risk of the coal samples considered in the study. Further, the partial dependence analysis was done to interpret the complex relationships between the WOP and intrinsic properties of coal.

Relationship of serum thyroid stimulating hormone with body mass index in healthy adults.

OBJECTIVE: To investigate any possible relationship between serum thyroid stimulating hormone (TSH) with body mass index (BMI) in healthy adults. MATERIALS AND METHODS: A total of 417 subjects aged 18-60 years who volunteered to get screened for thyroid illness with serum TSH have been enrolled from November 2012 to July 2013. Patients were divided into four groups based on BMI value: Underweight (BMI <18 kg/m(2)), normal (BMI: 18-22.9 kg/m(2)), overweight (BMI: 23-24.9 kg/m(2)), and obese (BMI ≥25 kg/m(2)). RESULT: In our study we found a significant variation (P < 0.001) in TSH with increasing BMI. As the BMI increased, mean TSH in the BMI range also increased. The individuals with higher BMI had higher TSH and this trend continued from underweight to Obese. The mean TSH of underweight group was 1.6036 mIU/L, normal weight group 2.1727 mIU/L, overweight group 2.2870 mIU/L and obese group 2.6416 mIU/L. CONCLUSION: In this study we found a significant relationship between serum TSH and BMI and mean TSH increased as BMI increased. Further large scale data from the population is required to confirm our findings.

Relevance of ankle reflex as a screening test for diabetic peripheral neuropathy.

BACKGROUND AND OBJECTIVE: The primary objective of this study is to find out if testing ankle reflex (AR) alone is an effective screening tool for diabetic peripheral neuropathy and the secondary objective of this study is to compare its sensitivity and specificity with biothesiometer vibration perception threshold (VPT), which is the gold standard. MATERIALS AND METHODS: A total of 450 patients with diabetes mellitus attending endocrine out-patient department from year 2012 January to 2013 July were included in this study. All patients underwent clinical assessment of AR compared with VPT by biothesiometer. RESULTS: AR is sensitive (81.09%) specific (81.679%) with diagnostic accuracy of (81.22%) and agreement between biothesiometer and AR is significant (κ = 0.538 P < 0.0001). CONCLUSION: AR is at par the gold standard that is biothesiometer VPT.