Description
Data Understanding
This assignment will require you to implement and interpret some of the data understanding concepts that were introduced in class, such as summary statistics and data visualization. Further,
you will be working with real-world data retrieved from an online repository, and while you will
be asked to utilize a variety of modules and functions, these have all been covered in the lecture
demos. Keep in mind that the main objective of this assignment is to highlight the insights that
we can derive from the data understanding process—the coding aspect is secondary. Accordingly, you are welcome to consult any online documentation and/or code that has been posted
to the course website, so long as all references and sources are properly cited. You are also encouraged to use code libraries, so long as you acknowledge any source code that was not written
by you by mentioning the original author(s) directly in your source code (comment or header).
You are expected to submit a single IPython Notebook file following the same
instructions and naming convention described in Assignment 0. Answers to the
conceptual questions can be embedded to the notebook file as markdown cells, and
you may use heading cells to further organize your document.
1 Iris Dataset (20 points)
Using your own module of choice (we recommend pandas), download the Iris flower dataset
available here1
into a DataFrame. For more details about the dataset and to obtain the feature
names, check this link. It is always recommended to familiarize yourself with the data you
intend to use for data mining purposes. The Iris dataset in particular has a rich history, having
been introduced in 1936 by Sir Ronald Fisher, often considered one of the fathers of modern
statistical theory.
1See the code we provided for Data Transformation here for an example of how this can be done.
1
1.1 Summary statistics
Print the first 5 elements of your DataFrame using the command head(). How many features
are there and what are their types (e.g., numeric, nominal)?
Compute and display summary statistics for each feature available in the dataset. These must
include the minimum value, maximum value, mean, range, standard deviation, variance,
count, and 25:50:75% percentiles.
1.2 Data Visualization
1.2.1 Histograms
To illustrate the feature distributions, create a histogram for each feature in the dataset. You
may plot each histogram individually or combine them all into a single plot. When generating
histograms for this assignment, use the default number of bins. Recall that a histogram
provides a graphical representation of the distribution of the data.
1.2.2 Boxplots
To further assess the data, create a boxplot for each feature in the dataset. All of the boxplots
will be combined into a single plot. Recall that a boxplot provides a graphical representation
of the location and variation of the data through their quartiles; they are especially useful for
comparing distributions and identifying outliers.
2 Pen-Based Handwritten Digits Dataset (20 points)
Repeat the same process described in Section 1 but this time load this dataset, which we
will refer to as Digits. Note that Digits is a much larger dataset than Iris, both with respect to
the number of instances and the number of features. A description of the dataset can be found
here.
3 Conceptual Questions (20 points)
Consider the histograms you generated for the Iris dataset. How do the shapes of the
histograms for petal length and petal width differ from those for sepal length and sepal width?
Now consider just the petal width histogram. Is there a particular value of petal length (which
ranges from 1.0 to 6.9) where the distribution of petal lengths (as illustrated by the
histogram) could be best segmented into two parts?
Now consider the boxplots you generated for the Iris dataset. There should be four boxplots,
one for each feature. Based upon these boxplots, is there a pair of features that appear to
have significantly different medians? Recall that the degree of overlap between variabilities is
an important initial indicator of the likelihood that differences in means or medians are
meaningful. Also, based solely upon the box plots, which feature appears to explain the
greatest amount of the data?
Lastly, consider the boxplots you generated for the Digits dataset. Do you observe any
outliers? If so, for what features? Now consider the corresponding histograms. What sort of
2
distribution do the second and forth features display? With that in mind, explain the outliers,
or lack thereof, in terms of what you observe from the histograms.
4 Graduate Student Portion (20 points / +10 points for
undergraduates)
Use random sampling with replacement to generate 5 new datasets that are each the same size
(same number of instances) as the original Iris dataset. Now compute and display the
summary statistics for each of these new datasets. Are they similar? Use at least one form of
visualization to contrast these datasets.
3
