Chapter 9: Data Modeling – Preprocessing

In this chapter, you will learn two important processes used to prepare data for modeling – splitting and scaling. You will learn how to use the sklearn methods – .StandardScaler and .MinMaxScaler for scaling, and .train_test_split for splitting. You will also be introduced to the reasons behind scaling and exactly what these methods do. As part of exploring splitting and scaling, you will use sklearn LinearRegression and statsmodels to create simple linear regression models.

By the end of this chapter, you will be comfortable preparing datasets to begin modeling. The main ideas you will learn in this chapter are as follows:

• Exploring independent and dependent variables
• Understanding data scaling and normalization
• Activity 9.01 – Data splitting, scaling, and modeling

An introduction to data modeling

Consider a statement such as the weather depends on the season. If we wanted to confirm this statement with data, we would collect information about the weather during different times of the year. The statement is asserting a model – a model of the weather that says we can say something about the weather if we know the season. Proposing and evaluating models is data modeling.

Often, we would like to understand the relationships within our data (numbers and other types of information), and in the previous chapter, we used visualization methods for that. Here, we can go a little deeper and ask questions such as are the independent variables correlated with each other? or is the output a linear function of the input? In some cases, we can answer these questions with charts; in other cases, we may construct a mathematical model. A mathematical model is simply a function that transfers some input data into an output.

Data modeling, the topic...

Exploring dependent and independent variables

In this chapter, you will learn about dependent and independent variables. You will learn about the need for scaling and normalization of data, in addition to performing those operations. You will also use some basic modeling methods to analyze your data.

At a high level, we can say a dependent variable is related to one or more independent variables in a linear or non-linear way. Linear models are easy to understand. A linear model relating one Y to one X is just a line. With multiple X variables, each one has a coefficient that gives its effect on Y, and since all those effects are independent, we just add all the effects together in a multivariate linear model. In a non-linear model, Y depends on X in a more complex way, such as Y being a function of X2. We can create non-linear models nearly as easily as linear models in pandas using some simple additional modules. We'll explore how to do that in the following chapter.

Much...

Understanding data scaling and normalization

If we inspect the coefficients of our mpg model, from the Training, validation, and test splits section earlier, we see that they range over several orders of magnitude. The code here iterates over the variable names and coefficient values, taking advantage of Python's .enumerate() method that iterates over the column names but also returns a counter, which we capture in coef and use to index the model coefficients. For reference, the code prints the range of the variable in data used to fit the model:

print('var\t  coef\t\t\t    range')

for coef, var in enumerate(my_data.columns[1:-1]):

    print(var, '\t', round(my_model.coef_[0][coef], 5), 

          '\twith range ', round(float(my_data[var].max() - 

               ...

Activity 9.01 – Data splitting, scaling, and modeling

You are charged with analyzing the performance of a combined cycle power plant and are given data on the full-load electrical power production along with environmental variables (such as temperature or humidity). In the first part of the activity, you will split the data manually and with sklearn, then you will scale the data, construct a simple linear model, and output the results:

1. For this activity, all you will need is the Pandas library, the modules from sklearn, and numpy. Load them in the first cell of the notebook.
2. Use the power_plant.csv dataset – 'Datasets\\power_plant.csv'. Read the data into a Pandas DataFrame, print out the shape, and list the first five rows.

The independent variables are as follows:

• AT – ambient temperature
• V – exhaust vacuum level
• AP – ambient pressure
• RH – relative humidity

The dependent variable is...

Summary

In this chapter, you learned how to split and scale data for downstream modeling tasks. You now can split data manually if that is appropriate but are also familiar with the sklearn methods to simplify the splitting tasks. You also saw how different scaling methods work and learned why min/max scaling might be used in some models and standardization in other models. You've seen how to make simple linear regression models, a topic to which we will return in the next chapter. Along the way, you learned why it is important to split data and hold some back from the modeling step in order to measure performance for new data. You now have the basic toolkit for preparing data for modeling, which is where we will begin the next chapter.