Redundant Questions Classification Based on Glove, XGBoost And Hyper Parameter Tuning |
Author(s): |
| Suraj Kumar Jalapally , Chaitanya Bharathi Institute of Technology; Subhasri Rallabandi, Chaitanya Bharathi Institute of Technology; Sacheth Reddy Mamidi, Chaitanya Bharathi Institute of Technology; Srikanth R, Chaitanya Bharathi Institute of Technology |
Keywords: |
| Quora Question Pair, XGBoost, Hyper Parameter Tuning |
Abstract |
|
A plethora of machine learning models have been put forth to model the sentence patterns accurately, which requires many features, parameters, and considerable computational ability. However, there was no extensive scrutiny in the number of relevant features, which compounds the model’s ability for classification. This paper has explored four different models to achieve a robust system to classify the duplicate question pair from the raw dataset published by Quora. All the challenges posed by the dataset, like data imbalance, unprocessed data, are addressed in this system. From the four models, the XGBoost model advocating different feature sets and Hyper parametric tuning helped us achieve the following best results: Accuracy of 84.2%, Precision 83.35%, Recall 82.75%, F1 score 83%. |
Other Details |
|
Paper ID: IJSRDV9I60168 Published in: Volume : 9, Issue : 6 Publication Date: 01/09/2021 Page(s): 339-342 |
Article Preview |
|
|
|
|
