CLASSIFICATION: LOGISTIC REGRESSION
This classification regression uses machine learning methodologies to predict a categorical outcome from numerical and categorical input parameters or predictors.For example, multiple laboratory flotation tests were completed on changing feed grade nickel ore, using different types of collectors. The results are trained/modelled to predict what is the best collector to use for a future ore at various feed grades.
To use:
- If you are making prediction from previously saved & uploaded model, skip to load model.
- The input data should be in a CSV file format with the predicted parameter at the last column and has values of 0 or 1.
- The columns containing numerical predictors should preceed the categorical predictors.
- Upload the training CSV data set to model.
- Change the model parameters where necessary.
- Train model. If the model is satisfactory, download (save) model for future prediction (saving 3 files).
- If predicting from the currently trained model, skip to upload test CSV file for prediction.
- To make prediction from saved/downloaded model, select the two model files (*.json and *.bin) and "Load Model". Then choose the metadata file and "Load Metadata".
- Upload upload test CSV file for prediction.
- Click on "Predict".
1. Upload Training CSV file
2. Train Model
Model Parameters:Depending on your data sets, modify the default model parameters below to fine tune the training until you are satisfied with the training results. Finding the right balance between these parameters is key to optimizing your model’s performance!
Learning Rate: Determines how much the model adjusts its weights in response to errors. A high learning rate makes fast updates but risks instability, while a low learning rate ensures gradual improvement but might take longer to converge.
Epochs: The number of times the model sees the entire dataset. More epochs allow more learning but can lead to overfitting if set too high.
Batch Size: Controls how many training samples are processed before updating the model’s weights. Smaller batches provide more updates and variance, while larger batches improve stability but need more memory.
and wait until the results appear below (this may take a while).
to be used for future predictions or on test data (csv file).
3. Load Model
If the model has been previously trained and downloaded/saved in files, upload both model files (*.json and *.bin).
Then upload the metadata file (*.json)
4. Upload Test CSV file for Prediction
Upload the test CSV file you want to make prediction for. The format of the test file should be the same as that of the training file used during model training.
Remember to train the model before making predictions or upload a saved model and its associated metadata file.