In this project, your aim is to leverage machine learning to automatically classify patients' examinations into five distinct classes using as predictors cardiac magnetic resonance radiomics.
Feature Engineering and Model Enhancement: Develop a pipeline that includes a dimensionality reduction or feature selection technique or a combination of both approaches as a prior step to classification.
Repeat the experiments conducted in step 2.
Try out at least:
two linear dimensionality reduction techniques
two nonlinear dimensionality reduction techniques
three feature selection techniques.
Provide details on the number of features used by each model, explaining the similarities and differences between the various approaches. Additionally, justify the concrete choices made in this process.
En 4. Feature Engineering and Model Enhancement -> a. Create different pipelines with different feature engineering strategies and models: falta escoger valores patron para el numero de dimensiones que queremos reducir el dataset. Como debemos escojer 4 tipos de dimensionality reduction, debemos elegir inicialmente 4 valores para n_components
En 4. Feature Engineering and Model Enhancement -> a. Create different pipelines with different feature engineering strategies and models: falta crear otros pipelines con otras combinaciones de feature selection, dim reduction y modelos
En 4. Feature Engineering and Model Enhancement -> b. Use the top 3 pipelines and perform hyperparameter tuning: a partir de los 3 mejores pipeline de a. optimizar los hyperparametros (mirar el ejemplo de las clases, pero pasar como parametro el splitter que configuramos antes o sea RandomizedSearchCV(..., cv=rskf_splitter))
En 5. Analysis: concluir cual el mejor modelo y escribir. Ya he escrito un poco acerca de lo que pensé - que nuestro enfoque es prediccion y no explability puesto que las features tampoco son simples de entender (son ya un resultado de preprocesamiento con pyradiomics)
Feature Engineering and Model Enhancement: Develop a pipeline that includes a dimensionality reduction or feature selection technique or a combination of both approaches as a prior step to classification.
Repeat the experiments conducted in step 2.
Try out at least:
Provide details on the number of features used by each model, explaining the similarities and differences between the various approaches. Additionally, justify the concrete choices made in this process.