Published Works (10)
Prediction of the compressive strength of normal concrete using ensemble machine learning approachβ
Contributors : Sanjog Chhetri Sapkota, Prasenjit Saha, Sourav Das, L. V. Prasad Meesaraganda
Highlights :
- Prediction of compressive strength based on laboratory data using ensemble bagging and boosting-based machine learning techniques
- Used new feature selection techniques and compared the performance of different combinations for optimization.
- This work showed better results than the traditional method, with an accuracy of 97 %.
Status : Published, Asian journal of civil engineering ( Springer, Scopus)
Link : https://link.springer.com/article/10.1007/s42107-023-00796-x
Modelling and validation of liquefaction potential index of fine-grained soils using ensemble learning paradigmsβ
Contributors : Sufyan Ghani, Sanjog Chhetri Sapkota, Raushan Kumar Singh, Abidhan Bardhan, Panagiotis G. Asteris
Highlights :
- Employed ensemble machine learning for the prediction of the liquefaction potential index of fine-grained soil.
- Higher prediction accuracy of 99%, which makes it feasible for early assessments of liquefaction susceptibility.
- verified using a validation set using a Chi-Chi earthquake for the developed model.
Status : Published in Soil Dynamics and Earthquake Engineering ( Elsevier, I.F 4)
A Generalised Explainable Approach to Predict the Hardened Properties of Self-Compacting Geopolymer Concrete Using Machine Learning Techniquesβ
Contributors : Endow Ayar Mazumder, Sanjog Chhetri Sapkota, Sourav Das, Prasenjit Saha, and Pijush Samui
Highlights :
- Employed hybrid machine learning to predict the hardened properties of self-compacting geopolymer concrete for the laboratory experiments .
- Higher prediction accuracy of more than 99% for compressive strength, flexural strength, and split tensile strength.
- Used shap explainable method for the model in giving the better decision with optimized results.
Status : Accepted (Yet to publish) in Computers and Structure (Techno press (SCI-indexed), I.F 4.1)
Optimized Machine Learning Models for Prediction of Effective Stiffness of Rectangular Reinforced Concrete Column Sectionsβ
Contributors : Sanjog Chhetri Sapkota, Sourav Das , Prasenjit Saha
Highlights :
- The primary objective of the present study is to predict the effective stiffness of rectangular RC columns using hybrid ensemble ML models combined with TPE-based Bayesian optimization.
- It was prepared by computing the effective stiffness ratio (π = πΈπΌπβπΈπΌπ) of columns from a set of RC frame buildings designed as per unified performance-based design (UPBD) methodology.Inputs include the amount of longitudinal rebar (ππ‘), the axial load (π), and the column section dimensions (π·π₯,π·π¦). Outputs are the effective stiffness ratios in both orthogonal directions (ππ₯, ππ¦).
- The results of the present study exhibit that the CatBoost model outperforms other considered machine learning models with π 2 value of 0.9921 and 0.9966.
- A sensitivity analysis based on SHAP has also been implemented to establish the correlation between input parameters and output parameters.
Status : Published in Structures, Elsevier, I.F 4.1
Link : https://doi.org/10.1016/j.istruc.2024.106155
PDF : https://drive.google.com/file/d/1vjm9KuIu0tHxjLpZ08754AGorgBh11rN/view
Discharge Determination of Rectangular Sharp Crested Weirs Using Machine Learning Modelsβ
Contributors : Sanjog Chhetri Sapkota, Mrimoy Dhar , Prasenjit Saha
Highlights :
- This paper highlights the use of hybrid ensemble ML algorithms and the comparison of standalone and ensemble algorithm.
- Highlights the factors affecting the coefficient of discharge with SHAP based Feature importance.
- Comparison of different existing empirical models and the better performing machine learning models.
- The additional insights on the variation of discharge coefficient with weir contraction ratio and Variation of discharge coefficient with the ratio h/P using machine learning results.
Status : Reviewing in (Flow Measurement and instrumentation, Elsevier, I.F 2.2)
Explainable Ensemble Soft Computing models for prediction of compressive strength of sustainable rice husk ash concreteβ
Contributors : Sanjog Chhetri Sapkota, Sagar Sapkota, Sani Isa aba , Gaurav Saini
Highlights :
- ML models are effective in predicting compressive strength.
- Efficient hybrid models with nested CV for prediction of strength were investigated.
- A comparison of models in terms of 12 performance indicators was carried out.
- SHAP analysis was used for local and global interpretation of the best model.
- The reverse design strategy for developing RHAC was analysed with influencing factors.
Status : Reviewing to Journal of Building Engineering ( Elsevier , IF : 6.9)
Explainable Ensemble Soft Computing models for prediction of compressive strength of sustainable rice husk ash concreteβ
Contributors : Sanjog Chhetri Sapkota, Sagar Sapkota, Sani Isa aba , Gaurav Saini
Highlights :
- ML models are effective in predicting compressive strength.
- Efficient hybrid models with nested CV for prediction of strength were investigated.
- A comparison of models in terms of 12 performance indicators was carried out.
- SHAP analysis was used for local and global interpretation of the best model.
- The reverse design strategy for developing RHAC was analysed with influencing factors.
Status : Reviewing to Journal of Building Engineering ( Elsevier , IF : 6.9)
Estimation of Reference Crop Evapotranspiration: A Comparative Analysisβ
Contributors : *Umesh Kumar Das ,Sanjog Chhetri Sapkota, Prasenjit Saha, Sameer Arora Highlights :
- Used several machine learning algorithms including standalone, bagging and boosting based ensemble for the prediction of the reference crop evapotranspiration in Jaipur, Rajasthan, India, which falls under arid regions of India.
- Higher prediction accuracy of more than 95% on peak seasons where evapotranspiration is seen higher.
- It captures the pattern of 100 years and the factors affecting evapotranspiration like precipitation, vapour pressure and average temperature.
Status : Reviewing in Hydrogeology journal , Springer (I.F 2.8)
Leveraging Explainable Hybrid Machine learning for the Prediction of Interfacial Bond Strength between Normal Concrete and Ultra-High-Performance Concreteβ
Contributors : Sanjog Chhetri Sapkota, Sagar Sapkota, Tushar Bansal Highlights :
- This paper highlights the use of hybrid algorithms (ensemble ML algorithms and TPE-based Bayesian optimization) with better predictability.
- Incorporation of a hybrid algorithm with Nested Cross validation for robust prediction in unseen data.
- Highlights the factors affecting the bond strength of UHPC-NSC, explaining local and global interperability using SHAP analysis.
- Addition of the reverse design method using Shap results for better insights into split tensile and slant shear data for UHPC-NSC bond strength.
Status : Reviewing in Structures, Elsevier, (I.F 4.1)
Prediction of Sustainable Concrete based on Palm Fibre and activated carbon using Boosting-Based based ensemble machine learning techniquesβ
Contributors : Sanjog Chhetri Sapkota, Musa Adamu , Prasenjit Saha, Sourav Das
Highlights :
- This paper highlights the use of Date Palm Fibre and activated carbon for as a sustainable materials.
- Use of hybrid (ensemble ML algorithm) and the comparison of standalone and ensemble algorithm.
- Highlights the factors affecting fresh and hardened properties using boosting based ensemble model.
- Comparison of different existing machine learning models.
- Shap Analysis like feature importance , dependence plots etc were used for model's explainability.
Status : Reviewing in Environmental Science and Pollution Research ,Springer, (I.F 5.8)
Behaviour of 3d printed concrete considering elevated temperatureβ
Contributors : Sanjog Chhetri Sapkota, Musa Adamu , Prasenjit Saha, Sourav Das
Highlights :
- This paper highlights the use of Date Palm Fibre and activated carbon for as a sustainable materials.
- Use of hybrid (ensemble ML algorithm) and the comparison of standalone and ensemble algorithm.
- Highlights the factors affecting fresh and hardened properties using boosting based ensemble model.
- Comparison of different existing machine learning models.
- Shap Analysis like feature importance , dependence plots etc were used for model's explainability.
Status : Reviewing in Environmental Science and Pollution Research ,Springer, (I.F 5.8)