Regularization adds a term to our loss function which penalizes large parameters. This can help prevent overfitting.

We introduce the simple linear regression model.

Adjusting our regression set up to accommodate categorical variables.

Pipelines are a nice way to put all modeling steps into one neat package. Here we introduce the most basic pipeline creation method.

A brief aside on how to handle time and dates in python.

Exponential Smoothing forecast methods are rolling averages with weights which decreases exponentially backwards through time. We introduce four such methods.

A Moving Average model expresses the value of a time series as a linear combination of lagged independent normally distributed errors.

We cover a plethora of data source websites you can use.

Your data is stuck in a database, can you get it out? Learn how in this video.

We take a moment to define clustering problems.

When you are missing data, try imputing!

Taking advantage of the gradient to minimize cost functions.

Plots related to threshold tuning, namely PR-curve and ROC-curve. The area under the ROC curve (AUC-ROC) measures the probability that class 1s are ranked as more probable than class 0s.

We discuss different loss functions (MSE, MAE, Huber) and evaluation metrics (MAPE, R^2, etc) for regression problems.

Scikit-learn models follow a simple model.fit(X,y), model.predict(X) API. This allows you to use many different models as black boxes so you can start using them right away.

A very short video which gives you pointers on where to find additional content on data collection. We do not cover this content in depth in the synchronous lectures.

We introduce the theory behind principal components analysis and demonstrate how it is implemented in python.

We review a typical workflow for a predictive modeling project with some baseball data.

We can make polynomial and nonlinear transformations of our features to make additional regression model types.

We introduce the notions of time series data and forecasting.

Some good baseline models for various flavors of time series data.

We learn what a stationary time series is and see how we can assess whether a time series is clearly non-stationary.

Where to go in order to keep learning about time series and forecasting.

We give a brief introduction into web scraping with BeautifulSoup

A cool PCA example where the components have an interesting interpretation.

Our first clustering algorithm.

We demonstrate some more advanced pipeline techniques in sklearn.

We introduce a sklearn object that makes hyper parameter tuning a tad easier.

Derivation of cross-entropy loss via negative log likelihood. Confusion matrix based classification metrics (accuracy, precision, recall, etc)

Train/test splits, cross-validation, and the frequent need for custom splits to give an "honest assessment" of model performance. Always think about what you want your model to generalize to!

Supervised learning models are trained on a dataset that includes both inputs (features) and the correct outputs (labels or target values). We discuss what they are and how they are trained.

Overview of the Data Science workflow (asking the right question, collecting/cleaning/exploring data, modeling, and deployment/reporting). A minimal example using concrete compressive strength.

Hyperparameter tuning and nested cross-validation.

We introduce multiple linear regression.

Having features with very different scales can impact the performance of some models. In this video we show how to use sklearn's StandardScaler object to standardize our features.

Since time series are ordered, we need a slight adjustment to our cross-validation strategy.

Our first time series forecast works by taking local averages of previous observations.

An autoregressive model expresses the value of a time series as a linear combination of lags plus an error term.

SARIMA stands for Seasonal Autoregressive Integrated Moving Average. We will learn how these models are defined and see an example of how to use such a model in practice.

How can we use python to collect data from APIs?

Another dimension reduction used primarily for data visualization.

Our second clustering algorithm.

All of your favorite classification algorithms back for regression purposes.

The expected generalization error of a learning algorithm can be decomposed into two terms: the bias and the variance.

Diagnostic plots for regression analysis, including residual vs. feature plots, residual vs. predicted value plots, and QQ-plots of residuals.

We often want to ingest raw data and apply a sequence of transformations (imputation, scaling, and feature engineering) to that data before modeling. Pipelines let us do that in a clean way.

Overview of the basics of data cleaning and exploratory data analysis. Showcases some fundamental tools (pandas, matplotlib, plotly) as well as some more specialized tools (panderas, missingno).