Pokemon Data Analysis Project

As part of Cambridge Spark's Level 4 Data Analysis Skills Bootcamp, we were required to create a portfolio piece showcasing our learning from the programme.

This repository serves as my portfolio piece looking closely at a large set of data. Obviously, I chose Pokemon data.

Read more about this repository on LinkedIn:

1. Part one
2. Part two

Data set

• The dataset is from kaggle
• It's pretty highly rated
• It's exhaustive in terms of Pokemon data

Data processing

Generally, the dataset was pretty clean and ready to work with.

I had to go through a few stages of processing and cleaning. Largely, this was removing columns that I didn't want to work with.

There was one column with a mistake in the name spelling, which I corrected.

Three columns had NaN values for which there were reasonable conversion to make.

Finally, there were a set of Pokemon with missing height and weight values. I looked up these missing values manually, plugged them into two separate lists and then merged those into the original dataset.

Analysing defence

I wanted to examine Pokemon statistics to look for interest. Perhaps correlation between type and stats. Or distributions of stats across generations.

I focused first on examining the defence stat.

Determining the mean, median and modal values for this stat. They are:

{
    mean: 73,
    median: 70,
    mode: 50
}

Then I mapped a bar chart of all defence stats.

There seemed to be a trend emerging from this graph. However, values occuring only a few times skews the data's representation. By examining the stat in intervals, we can see more distinctly that there is a clear trend around the lower end of the scale which gradually levels out as the stat peaks.

Some interesting points here:

• Left to right can help us examine the values across generations
• The highest 3 values appear to be two steel types and a bug type
• Next we start to see appearances of a few rock, ground, and water types with steel and bug appearing again

We have a few outliers to the set of types we've seen already as we get further down. Examining these will be interesting. Looking at the graph, I can see that these outliers sit in two possible sets:

• Pokemon with a defence value higher than 125
• Top 25 Pokemon by defence

So let's examine these.

Pokemon with a defence value higher than 125

1	2	3	4	5
Golem	Slowbro	Cloyster	Onix	Forretress
Steelix	Scizor	Shuckle	Skarmory	Tyranitar
Lugia	Nosepass	Lairon	Aggron	Torkoal
Dusclops	Relicanth	Salamence	Metagross	Regirock
Registeel	Groudon	Bastiodon	Rhyperior	Leafeon
Probopass	Dusknoir	Uxie	Gigalith	Audino
Cofagrigus	Carracosta	Ferrothorn	Cobalion	Doublade
Carbink	Avalugg	Wishiwashi	Toxapex	Golisopod
Pyukumuku	Turtonator	Cosmoem	Buzzwole	Kartana

List of types, no leg: 'rock' 'water' 'bug' 'steel' 'fire' 'ghost' 'dragon' 'ground' 'grass' 'normal' 'ice' 'poison'