if someone help me out to improve the performace of the model then that would be great help, i am using ExtraTreeClassifier with 8 folds, but i want to imrove the performace of the model, if some one guides me in the Feature Engineering part of it or modellling part that helps a lot, Thanks in Advance.
", "messages": [ { "id": "2247596", "postDate": "05/06/2023 06:24:58", "content": "if someone help me out to improve the performace of the model then that would be great help, i am using ExtraTreeClassifier with 8 folds, but i want to imrove the performace of the model, if some one guides me in the Feature Engineering part of it or modellling part that helps a lot, Thanks in Advance.
", "rawMarkdown": "if someone help me out to improve the performace of the model then that would be great help, i am using ExtraTreeClassifier with 8 folds, but i want to imrove the performace of the model, if some one guides me in the Feature Engineering part of it or modellling part that helps a lot, Thanks in Advance.", "votes": null }, { "id": "2247735", "postDate": "05/06/2023 08:40:31", "content": "Did you try GridSearchCV, for hyper parameter tuning? Try also several folds (10,20) as well as other Classifiers.
", "rawMarkdown": "Did you try GridSearchCV, for hyper parameter tuning? Try also several folds (10,20) as well as other Classifiers.", "votes": null }, { "id": "2253584", "postDate": "05/10/2023 09:50:44", "content": "I want to understand more in terms of feature engineering, but this is also helpful
", "rawMarkdown": "I want to understand more in terms of feature engineering, but this is also helpful", "votes": null }, { "id": "2253598", "postDate": "05/10/2023 10:14:07", "content": "You can iterate and experiment with the following different approaches to find the best combination for your particular scenario. Good luck!
\nFeature Scaling: Ensure that your features are on a similar scale. Scaling can help improve the performance of certain models, especially those sensitive to feature magnitudes. Common scaling techniques include standardization (mean=0, standard deviation=1) or normalization (scaling to a range between 0 and 1).
\nFeature Transformation: Consider transforming your features to capture non-linear relationships or reduce skewness in the data. Some common transformations include logarithmic, square root, or Box-Cox transformations.
\nEncoding Categorical Variables: If your dataset contains categorical variables, you need to encode them numerically. One-hot encoding and label encoding are commonly used techniques. Additionally, you can explore target encoding, which encodes categories based on their target variable statistics.
\nHandling Missing Values: Deal with missing values appropriately. Depending on the amount and nature of missing data, you can either remove instances or impute missing values using techniques such as mean imputation, median imputation, or advanced methods like K-nearest neighbors or regression imputation.
\nFeature Generation: Sometimes, creating new features based on domain knowledge or interactions between existing features can provide valuable information. For example, you can derive features by combining or interacting existing features or extracting meaningful information from date/time variables.
\nRegularization: Regularization techniques like L1 (LASSO) or L2 (Ridge) regularization can help reduce overfitting and improve model generalization. They penalize the magnitude of feature coefficients, encouraging simpler and more robust models.
\nEnsemble Methods: Consider using ensemble methods like Random Forests or Gradient Boosting. These models combine multiple weak learners to create a stronger predictor. They often perform well and handle complex interactions between features effectively.
\nHyperparameter Tuning:Optimize the hyperparameters of your model to find the best combination for your specific dataset. You can use techniques like grid search, random search, or Bayesian optimization to search for the optimal hyperparameter values.
\nCross-Validation: Make sure you're using appropriate cross-validation techniques. Instead of simply using 8-fold cross-validation, consider using techniques like stratified k-fold or nested cross-validation to obtain more reliable estimates of model performance.
", "rawMarkdown": "You can iterate and experiment with the following different approaches to find the best combination for your particular scenario. Good luck!\n\n**Feature Scaling:** Ensure that your features are on a similar scale. Scaling can help improve the performance of certain models, especially those sensitive to feature magnitudes. Common scaling techniques include standardization (mean=0, standard deviation=1) or normalization (scaling to a range between 0 and 1).\n\n**Feature Transformation:** Consider transforming your features to capture non-linear relationships or reduce skewness in the data. Some common transformations include logarithmic, square root, or Box-Cox transformations.\n\n**Encoding Categorical Variables:** If your dataset contains categorical variables, you need to encode them numerically. One-hot encoding and label encoding are commonly used techniques. Additionally, you can explore target encoding, which encodes categories based on their target variable statistics.\n\n**Handling Missing Values:** Deal with missing values appropriately. Depending on the amount and nature of missing data, you can either remove instances or impute missing values using techniques such as mean imputation, median imputation, or advanced methods like K-nearest neighbors or regression imputation.\n\n**Feature Generation:** Sometimes, creating new features based on domain knowledge or interactions between existing features can provide valuable information. For example, you can derive features by combining or interacting existing features or extracting meaningful information from date/time variables.\n\n**Regularization:** Regularization techniques like L1 (LASSO) or L2 (Ridge) regularization can help reduce overfitting and improve model generalization. They penalize the magnitude of feature coefficients, encouraging simpler and more robust models.\n\n**Ensemble Methods:** Consider using ensemble methods like Random Forests or Gradient Boosting. These models combine multiple weak learners to create a stronger predictor. They often perform well and handle complex interactions between features effectively.\n\n**Hyperparameter Tuning:**Optimize the hyperparameters of your model to find the best combination for your specific dataset. You can use techniques like grid search, random search, or Bayesian optimization to search for the optimal hyperparameter values.\n\n**Cross-Validation:** Make sure you're using appropriate cross-validation techniques. Instead of simply using 8-fold cross-validation, consider using techniques like stratified k-fold or nested cross-validation to obtain more reliable estimates of model performance.", "votes": null }, { "id": "2253656", "postDate": "05/10/2023 11:31:21", "content": "Feature generation is very tricky part, and one of the most important part, if you put some lights on this topic it will heps a lot to me and others as well.
", "rawMarkdown": "Feature generation is very tricky part, and one of the most important part, if you put some lights on this topic it will heps a lot to me and others as well.", "votes": null } ], "comments": [ { "id": 2247735, "author_name": "catadanna", "author_url": "", "post_date": "05/06/2023 08:40:31", "content": "Did you try GridSearchCV, for hyper parameter tuning? Try also several folds (10,20) as well as other Classifiers.
", "votes": null, "replies": [ { "id": 2253584, "author_name": "mfumar6", "author_url": "", "post_date": "05/10/2023 09:50:44", "content": "I want to understand more in terms of feature engineering, but this is also helpful
", "votes": null, "replies": [] } ] }, { "id": 2253598, "author_name": "pranayakuniyal", "author_url": "", "post_date": "05/10/2023 10:14:07", "content": "You can iterate and experiment with the following different approaches to find the best combination for your particular scenario. Good luck!
\nFeature Scaling: Ensure that your features are on a similar scale. Scaling can help improve the performance of certain models, especially those sensitive to feature magnitudes. Common scaling techniques include standardization (mean=0, standard deviation=1) or normalization (scaling to a range between 0 and 1).
\nFeature Transformation: Consider transforming your features to capture non-linear relationships or reduce skewness in the data. Some common transformations include logarithmic, square root, or Box-Cox transformations.
\nEncoding Categorical Variables: If your dataset contains categorical variables, you need to encode them numerically. One-hot encoding and label encoding are commonly used techniques. Additionally, you can explore target encoding, which encodes categories based on their target variable statistics.
\nHandling Missing Values: Deal with missing values appropriately. Depending on the amount and nature of missing data, you can either remove instances or impute missing values using techniques such as mean imputation, median imputation, or advanced methods like K-nearest neighbors or regression imputation.
\nFeature Generation: Sometimes, creating new features based on domain knowledge or interactions between existing features can provide valuable information. For example, you can derive features by combining or interacting existing features or extracting meaningful information from date/time variables.
\nRegularization: Regularization techniques like L1 (LASSO) or L2 (Ridge) regularization can help reduce overfitting and improve model generalization. They penalize the magnitude of feature coefficients, encouraging simpler and more robust models.
\nEnsemble Methods: Consider using ensemble methods like Random Forests or Gradient Boosting. These models combine multiple weak learners to create a stronger predictor. They often perform well and handle complex interactions between features effectively.
\nHyperparameter Tuning:Optimize the hyperparameters of your model to find the best combination for your specific dataset. You can use techniques like grid search, random search, or Bayesian optimization to search for the optimal hyperparameter values.
\nCross-Validation: Make sure you're using appropriate cross-validation techniques. Instead of simply using 8-fold cross-validation, consider using techniques like stratified k-fold or nested cross-validation to obtain more reliable estimates of model performance.
", "votes": null, "replies": [ { "id": 2253656, "author_name": "mfumar6", "author_url": "", "post_date": "05/10/2023 11:31:21", "content": "Feature generation is very tricky part, and one of the most important part, if you put some lights on this topic it will heps a lot to me and others as well.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "2247596": "if someone help me out to improve the performace of the model then that would be great help, i am using ExtraTreeClassifier with 8 folds, but i want to imrove the performace of the model, if some one guides me in the Feature Engineering part of it or modellling part that helps a lot, Thanks in Advance.", "2247735": "Did you try GridSearchCV, for hyper parameter tuning? Try also several folds (10,20) as well as other Classifiers.", "2253584": "I want to understand more in terms of feature engineering, but this is also helpful", "2253598": "You can iterate and experiment with the following different approaches to find the best combination for your particular scenario. Good luck!\n\n**Feature Scaling:** Ensure that your features are on a similar scale. Scaling can help improve the performance of certain models, especially those sensitive to feature magnitudes. Common scaling techniques include standardization (mean=0, standard deviation=1) or normalization (scaling to a range between 0 and 1).\n\n**Feature Transformation:** Consider transforming your features to capture non-linear relationships or reduce skewness in the data. Some common transformations include logarithmic, square root, or Box-Cox transformations.\n\n**Encoding Categorical Variables:** If your dataset contains categorical variables, you need to encode them numerically. One-hot encoding and label encoding are commonly used techniques. Additionally, you can explore target encoding, which encodes categories based on their target variable statistics.\n\n**Handling Missing Values:** Deal with missing values appropriately. Depending on the amount and nature of missing data, you can either remove instances or impute missing values using techniques such as mean imputation, median imputation, or advanced methods like K-nearest neighbors or regression imputation.\n\n**Feature Generation:** Sometimes, creating new features based on domain knowledge or interactions between existing features can provide valuable information. For example, you can derive features by combining or interacting existing features or extracting meaningful information from date/time variables.\n\n**Regularization:** Regularization techniques like L1 (LASSO) or L2 (Ridge) regularization can help reduce overfitting and improve model generalization. They penalize the magnitude of feature coefficients, encouraging simpler and more robust models.\n\n**Ensemble Methods:** Consider using ensemble methods like Random Forests or Gradient Boosting. These models combine multiple weak learners to create a stronger predictor. They often perform well and handle complex interactions between features effectively.\n\n**Hyperparameter Tuning:**Optimize the hyperparameters of your model to find the best combination for your specific dataset. You can use techniques like grid search, random search, or Bayesian optimization to search for the optimal hyperparameter values.\n\n**Cross-Validation:** Make sure you're using appropriate cross-validation techniques. Instead of simply using 8-fold cross-validation, consider using techniques like stratified k-fold or nested cross-validation to obtain more reliable estimates of model performance.", "2253656": "Feature generation is very tricky part, and one of the most important part, if you put some lights on this topic it will heps a lot to me and others as well." }, "source": "meta" }