Scenario

Kaggle contains a variety of datasets for experimentation. One such dataset is the $80-$cereals dataset. This dataset contains the consumer reports rating of several different cereals along with various data about the cereals. The dataset contains the following fields.

Content

Fields in the Dataset

Name: Name of cereal

mfr: Manufacturer of cereal

A $=$ American Home Food Products

G $=$ General Mills

K $=$ Kellogg$$s

N $=$ Nabisco

P $=$ Post

Q $=$ Quaker Oats

R $=$ Ralston Purina

Type:

Cold

Hot

Calories: Calories per serving

Protein: Grams of protein

Fat: Grams of fat

Sodium: Grams of sodium $($Note: The original data contains milligrams.$)$

Fiber: Grams of dietary fiber

Carbs: Grams of complex carbohydrates

Sugars: Grams of sugars

Potassium: Grams of potassium $($Note: The original data contains milligrams.$)$

Vitamins: Vitamins and minerals: $0,25,$ or $100,$ indicating the typical percentage of FDA recommended

Shelf: Display shelf $(1,2,$ or $3,$ counting from the floor$)$

Weight: Weight in ounces of one serving

Cups: Number of cups in one serving

Rating: A rating of the cereals

Create a new experiment in Azure Machine Learning and perform linear regression on the cereals.csv dataset. We are seeking to determine if we can predict the rating based on the above features. The ratings are based on nutrition and taste. Open the cereals.csv file and look through the data to familiarize yourself with the data. You will need to cleanse the data$$are there any missing values? Any outliers? Do you need to make any fields categorical? You will also need to normalize the data. You will want to develop python visualizations $($feel free to use the code we have used in previous examples to create a pairwise scatter plot$).$ You may also want to focus on a few selected fields based on your visualizations $($sugar$,$ calories, fat, fiber, etc$).$ Don$$t forget to split the data into a training and test set. You will need to create a linear regression model, train the model, score the model, and then evaluate it$.$

How accurate is your model? What is the R$2$ value?

Look at the python pairwise scatter plot. What features do you want to remove from the model? Continue iterating $($removing features and adding them back in$)$ and reviewing the evaluation results. Find the best R$2$ value you can. Try to get above $0\mathrm{.}70.$ What features are good predictors of the cereal rating?

Prepare a report on your regression analysis. In your report include several screenshots and explanations of your model. The rubric is below.

Introduction to the problem

Uploading dataset

Data preparation $($normalization$)$

Data Visualization $($python script$)$

Selecting features

Splitting data

Linear regression model

Training the model

Scoring the model $$ show scored labels

Evaluating the model

Iteration process

Why you chose to exclude certain features

New evaluation when those features were excluded

What features are most influential on the rating

Conclusion

Challenges

Career skills obtained

Web service $($optional$)$

Note: Pairwise scatter plot code in Python.

def azureml$\_$main$($frame$1)$:

## import libraries

import matplotlib

matplotlib.use$(\text{'}$agg$\text{'})$ # Set backend

from pandas.tools.plotting import scatter$\_$matrix

import pandas.tools.rplot as rplot

import matplotlib.pyplot as plt

import numpy as np

## Create a pair$-$wise scatter plot

Azure $=$ True

fig$1=$ plt$.$figure$(1,$ figsize$=(10,10))$

ax $=$ fig$1.$gca$()$

sm$=$scatter$\_$matrix$($frame$1,$ alpha$=0\mathrm{.}3,$

diagonal$=$'kde', ax $=$ ax$)$

$[$s$.$xaxis.label.set$\_$rotation$(45)$ for s in sm$.$reshape$(-1)]$

$[$s$.$yaxis.label.set$\_$rotation$(45)$ for s in sm$.$reshape$(-1)]$

plt$.$show$()$

fig$1.$savefig$(\text{'}$scatter$1.$png$\text{'})$