Video exercises: Slides & transcript

Course Creation at DataCamp

Kelsey McNeillie

Instructor Success Manager

Video exercises

First part of a lesson
Learning objective
Presentation

add video exercise.png

Pick a slide type

Structured scripts

Exact words recorded
What → Why → How
Sentences should flow naturally
Be brief
- Ideally 400-500 words

Build a narrative

Storytelling
Be ADEPT
1. Use an Analogy
2. Draw a Diagram
3. Provide a concrete Example
4. Describe it in Plain English
5. Provide a Technical definition

Timeless scripts

Don't	Do
Manipulating Time Series Data in Python	course on time series manipulation in Python
I'm a Data Scientist at Data Company X	I'm a Data Scientist
I'm writing a book on Machine Learning	Follow me on Twitter for updates
The current version of this package...	<Focus on the utility, not the implementation>

Dynamic slides

Dynamic:
- Add animations to lists
- Add animations to code
Avoid "deadtime"
- Add movement (at least) every 30 seconds
Animate bullet points

Bullet points: What you should NOT do

Machine learning workflow

Data exploration
Data processing
Modeling
Evaluation
Improvements

Bullet points: What you should do

Machine Learning workflow

Data exploration {{1}} →
Data processing {{2}} →
Modeling {{3}} →
Evaluation {{4}} →
Improvements {{5}} →

Machine Learning workflow

Data exploration
Data processing
Modeling
Evaluation
Improvements

Code: What you should NOT do

# Define trainControl function
fitControl <- trainControl(method = "adaptive_cv", number = 3, repeats = 3,
                           adaptive = list(min = 3, alpha = 0.05, method = "BT", complete = FALSE),
                           search = "random")
# Start timer
tic()

# Train model
svm_model_voters_ar <- train(turnout16_2016 ~ ., data = voters_train_data, 
                             method = "svmPoly", trControl = fitControl,
                             verbose = FALSE, tuneLength = 6)

# Stop timer
toc()

1.675 sec elapsed

Code: What you should do

# Define trainControl function
fitControl <- trainControl(method = "adaptive_cv", number = 3, repeats = 3,
                           adaptive = list(min = 3, alpha = 0.05, method = "BT", complete = FALSE),
                           search = "random")

# Start timer
tic()

# Train model
svm_model_voters_ar <- train(turnout16_2016 ~ ., data = voters_train_data, 
                             method = "svmPoly", trControl = fitControl,
                             verbose = FALSE, tuneLength = 6)

# Stop timer
toc()