Data Analysis From Scratch With Python: Step By Step Guide

print(pred) #predicted labels i.e flower species
print(y_test) #actual labels
print((accuracy_score(pred, y_test)))*100 #prediction accuracy
#Reference: 
http://docs.python-
guide.org/en/latest/scenarios/ml/
If we run the code, we’ll get something like this: 
[0 1 1 1 0 2 0 2 2 2]
[0 1 1 1 0 2 0 2 2 2]
100.0
The first line contains the predictions (0 is Iris setosa, 1 is Iris versicolor, 2 is Iris
virginica). The second line contains the actual flower species as indicated in the
dataset. Notice the prediction accuracy is 100%, which means we correctly
predicted each flower’s species.
These might all seem confusing at first. What you need to understand is that the
goal here is to create a model that predicts a flower’s species. To do that, we split
the data into training and test sets. We run the algorithm on the training set and
use it against the test set to know the accuracy. The result is we’re able to predict
the flower’s species on the test set based on what the computer learned from the
training set.
Potential & Implications
It’s a quick and simple example. But its potential and implications can be
enormous. With just a few modifications, you can apply the workflow to a wide
variety of tasks and problems.
For instance, we might be able to apply the same methodology on other flower
species, plants, and animals. We can also apply this in other Classification
problems (more on this later) such as determining if a cancer is benign or
malignant, if a person is a very likely customer, or if there’s a human face in the
photo.
The challenge here is to get enough quality data so our computer can properly
get “good training.” It’s a common methodology to first learn from the training
set and then apply the learning into the test set and possibly new data in the
future (this is the essence of machine learning).
It’s obvious now why many people are hyped about the true potential of data
analysis and machine learning. With enough data, we can create automated

---

systems on predicting events and classifying objects. With enough X-ray images
with corrects labels (with lung cancer or not), our computers can learn from the
data and make instant classification of a new unlabeled X-ray image. We can
also apply a similar approach to other medical diagnosis and related fields.
Back then, data analysis is widely used for studying the past and preparing
reports. But now, it can be used instantaneously to predict outcomes in real time.
This is the true power of data, wherein we can use it to make quick and smart
decisions.
Many experts agree that we’re just still scratching the surface of the power of
performing data analysis using significantly large datasets. In the years to come,
we’ll be able to encounter applications never been thought before. Many tools
and approaches will also become obsolete as a result of these changes.
But many things will remain the same and the principles will always be there.
That’s why in the following chapters, we’ll focus more on getting into the
mindset of a savvy data analyst. We’ll explore some approaches in doing things
but these will only be used to illustrate timeless and important points.
For example, the general workflow and process in data analysis involve these
things: ● Identifying the problem (asking the right questions) ● Getting &
processing data ● Visualizing data ● Choosing an approach and algorithm ●
Evaluating the output ● Trying other approaches & comparing the results ●
Knowing if the results are good enough (knowing when to stop) It’s good to
determine the objective of the project first so we can set clear expectations and
boundaries on our project. Second, let’s then gather data (or get access to it) so
we can start the proper analysis. Let’s do that in the next chapter.

---