K-Nearest Neighbors implementation in R with encoding and feature scaling.

Quick implementation of XGBoost to compare to implementation of random forest algorithm against same data; hint XGBoost is much faster both in coding and in execution (no EPOCH's!)

Example of using Grid Search to optimize parameters of a model to build a model with optimal parameters. As compared to our specifying the parameters in our implementation.

Using K-Fold Cross Validation to improve the accuracy of our model.

Applying Kernel PCA to data before we then apply a linear model. This application is called the Kernel Trick.

Like PCA LDA in a nutshell; what we are doing is taking a large number of independent variables and extracting them down to a core group of new independent variables that best describe the relationship (most variance) of the data in the dataset. A distinction of LDA is that this extraction is done WITH knowledge of the dependent variable so the LDA process is considered a supervised dimensionality reduction technique.

PCA in a nutshell; what we are doing is taking a large number of independent variables and extracting them down to a core group of new independent variables that best describe the relationship (most variance) of the data in the dataset. Because this extraction is done without knowledge of the dependent variable the PCA process is considered un-supervised dimensionality reduction technique.

Implementation of an Artificial Neural Network in R using the h2o package.

Implement a Random Forest Algorithm with a sparse matrix set of data to identify positive reviews from negative reviews.

Implementation from scratch a Thompson Sampling algorithm to build a machine learning model based on a multi-arm bandit problem.

In this example a UCB is built in a loop and applied to data to modify which ad is being used based on how well the ads are doing in real life data.

Investigation of a dataset using Association Learning Algorithm - Eclat to determine common sets of items occurring in the data based on a given level of support.

Investigation of a dataset using Association Learning Algorithm - Apriori to determine rules based on support in the dataset, a level of confidence and the lift associated with each.

Using Hierarchical Clustering to create a Dendogram to determine optimal clustering from which then build a clustering model.

Using methodology of the Elbow Method, using KMeans Clustering to then use KMeans Clustering to work out the best classification model for our data.

Some things I've run across that seem useful to save separate from where I've used them.

Use of a random forest prediction with 500 trees in R.

Using a Decision Tree algorithm (rpart) to separate the data and build a model to predict outcomes.

Create prediction model in R with the Naive Bayes algorithm.

Example of FlexDashboards with tabs, displaying various graphs. Very cool!

Very cool R markdown code for creating dashboards and presenting data.

Using various SVM Kernels in R to construct a machine and fit to a dataset. In this example we use the radial and linear to compare their ability to model non-linearly separable data. No sunrise the radial or Gaussian model is better.

Build a SVM machine in R, fit to data and visualize predictions.

An example of the moving average crossover trading system with some playing with the in and out trade notifiers based on different trends in the market.

An example of the moving average crossover trading system.

Basics of pulling stock market data; Open, High, Low, Close, Volume and Adjusted and then plotting and analyzing with the quantmod package.