Ensembling and stacking are made to improve the model's predictive capacity. But your need here is to enable the model to predict a new Class.
This doesn't happen in machine Learning unless you use your domain knowledge and infer that from the individual prediction.
e.g. If I train a model to predict white color and another to train Black color, it doesn't ...
I will say, it's an either Or situation
You can pick one of "Incremental/Online" training Or "addition of new class".
You may do a fine-tuning approach with a Neural network by adjusting the o/p layer and training the last few layers. But this approach expects the new data to be quite similar to the training set.
KNN - Can do the online ...
There are something like 30 random forest packages in R. "randomForest" is one of the first implementations and so is well known, but it's not great for large datasets. "ranger" is a good R package; it's fast, handles large data, and has parameter tuning searches. It's easier to use with package "parsnip".
It is difficult to answer your question without the access to your code. The best way to understand the difference is to profile the code and see where the bottlenecks are for your specific problem.
For this, you can use different profiling modules in python:
python line profiler
You can do a K Means Clustering to see cluster your products and see if some products is situated very closely. (In the same cluster). Then you can say that products in the same cluster are similar. But you have to find the optimal k value of clusters.
The fmla is the model formula: fmla <- formula(blood_pressure ~ age + weight)
So the correct solution should be
# bloodpressure is in the workspace
# Create the formula and print it
fmla <- formula(blood_pressure ~ age + weight)
# Fit the model: bloodpressure_model
bloodpressure_model <- lm(fmla, data = bloodpressure)
This is an implementation detail, and I wouldn't necessarily rely on this behavior, but presently in sklearn, it will choose the "first" class.
The predict method calls for the probability prediction, then takes the argmax, which in case of ties takes the first one:
First, it would be beneficial if you could mention whether that $R2$ presented in the figure is for the test or training dataset.
Let's assume that it is for the test dataset.
The answer to all of your questions is subjective and depends on the type of data you used for this purpose. But let me briefly answer them.
Because the Random Forest model (RF) is an ...
It should be a bug in their server. The fmla variable should have the same contents in your code and in theirs. This is because the last assignment on both scripts is
fmla <- lm(blood_pressure ~ age + weight, data=bloodpressure)
You can do mainly two things: bootstrap or oversampling.
With statistical data you can do bootstrapping (random sampling with replacement)
Bagging methods help boosting you model accuracy. The pseudocode will be a bit like this.
for estimator in range(number of estimators):
Sampling some data
Fitting a model
This way ...
For a good overview of common statistical methods, check our Rob Hyndman's Forecasting Principles and Practices.
For a comparison of different models, including machine learning and deep learning, check out the M4 competition and their companion repo. The current state-of-the-art on the M4 data set is N-BEATS as far as I know.
For a replication and ...
This error usually indicates your train and test in the train_test_split() function call are different sizes. You may need to reshape one to get them to match. Look at the shapes of train and test to see what is the problem.
If one understands the business problem you are trying to solve, then this shouldn't be that hard.
You can present the results of your model. Understand your data first, is it skewed, balanced? Understand your model, check the feature importance. Do they make sense with the domain knowledge you have? For example, if your data is highly imbalanced and you ...
You may be able to convert the entire model pipeline to a standardized format. PMML is such a format, and there are tools (e.g. jpmml) to convert all your named modeling package objects to PMML, though perhaps you've used something else that isn't already easily-converted.
Otherwise, just force installation of dependencies (and make it easy), through a ...
You can try to fuse the image and audio data. You can fuse in different ways:
1) Early fusion - fusion at feature level
2) Late fusion - fusion at output level (similar to your idea)
Here is some papers for your reference:
Look, Listen and Learn
Learn to Combine Modalities in Multimodal Deep