A friendly tutorial on getting zip codes and other geographic data from street addresses.
Knowing how to deal with geographic data is a must-have for a data scientist. In this post, we will play around with the MapQuest Search API to get zip codes from street addresses along with their corresponding latitude and longitude to boot!
In 2019, my friends and I participated in CivTechSA Datathon. At one point in the competition, we wanted to visualize the data points and overlay them on San Antonio’s map. The problem is, we had incomplete data. Surprise! All we had were a street number and a street name — no zip code, no latitude, nor longitude. We then turned to the great internet for some help.
We found a great API by MapQuest that will give us exactly what we needed. With just a sprinkle of Python code, we were able to accomplish our goal.
Today, we’re going to walk through this process.
To follow along, you can download the data from here. Just scroll down to the bottom tab on over to the Data Catalog 2019. Look for SAWS (San Antonio Water System) as shown below.
Download the file by clicking on the link to the Excel file.
If you’re using Windows 10, it will ask you to open Microsoft Store.
Go ahead and click on the “Install” button.
And let’s get started by clicking on the “Launch” button.
A Thousand Clicks
Click on “Get data” when the splash screen appears.
You will be presented with a lot of file format and sources; let’s choose “Text/CSV” and click on the “Connect” button.
Select “order_products_prior.csv” and click on the “Open” button.
The image below shows what the data looks like. Click on the “Load” button to load the dataset into Power BI Desktop.
Load the rest of the dataset by selecting “Get Data” and choosing the “Text/CSV” option on the dropdown.
You should have these three files loaded into Power BI Desktop:
You should see the following tables appear on the “Fields” panel of Power BI Desktop, as shown below. (Note: the image shows Power BI in Report View.)
Let’s see what the Data View looks like by clicking on the second icon on the left side of Power BI Desktop.
And now, let’s check out the Model View where we will see how the different tables are related to each other.
If we hover a line, it will turn yellow and the corresponding related fields are both highlighted as well.
In this case, Power BI Desktop is smart to infer the two relationships. However, most of the time, we will have to create the relationships ourselves. We will cover this topic in the future.
Let’s go back to the Report View and examine the “Visualizations” panel closely. Look for the “slicer” icon which looks like a square with a funnel at the bottom right corner. Click on it to add a visual to the report.
In the “Fields” panel, find the “department_id” and click the checkbox on its left.
This will cause the “department_id” field to appear under the “Visualizations” panel in the “Field” box.
Next, take your mouse cursor and hover over the top right corner of the visual in the Report View. Click on the three dots that appeared in the corner as shown below.
Click on “List” in the dropdown that appeared.
While the “department_id” visual is selected, you should see corner marks indicating the visual as the active visual. While the “department_id” is active, press CTRL+C to copy it and then CTRL+V to paste it. Move the new visual to the right of the original visual.
Make the second visual active by clicking somewhere inside it. Then look for the “aisle_id” field in the “Fields” panel on the right of Power BI Desktop as shown below.
Try selecting a value on the “department_id” visual and observe how the selection on “aisle_id” changes accordingly.
Now, examine the “Visualizations” panel again and click on the table visual as shown below.
In the “Fields” panel, select “product_id” and “product_name” or drag them in the “Values” box.
Power BI Desktop should look similar to the image below.
This time, try selecting a value from both “department_id” and “aisle_id” — observe what happens to the table visual on the right.
Let’s create another visual by copying and pasting the table visual. This time, select (or drag) the following fields to the “Values” box of the visual.
Power BI Desktop should now look similar to the image below.
Try clicking one of the selections in the table visual (where it’s showing “product_id” and “product_name”) and observe how the table on the right changes accordingly.
For a closer look, activate Focus Mode by clicking on the icon as shown below.
The table displays the details of orders that have the product that you selected in the table with “product_id” and “product_name.”
Get out of Focus Mode by clicking on “Back to report” as shown below.
Let’s rename this page or tab by right-clicking on the page name (“Page 1”) and selecting “Rename Page.”
Type in “PRODUCTS” and press ENTER.
let’s add another page or tab to the report by right-clicking on the page name again (“PRODUCTS”) and selecting “Duplicate Page.”
Rename the new page “TRANSACTIONS” and delete (or remove) the right-most table with order details on it.
Change the top-left visual and make update the fields as shown below. The “Fields” box should say “order_dow” while the top-left visual is activated.
Move the visuals around so it looks similar below.
Do the same thing for the next visual. This time, select “order_hour_of_day” and your Power BI Desktop should like the image below.
Do the same thing one last time for the last table and it should now contain fields as shown below.
Let’s add another page or tab to the report by clicking on the “+” icon at the bottom of the report’s main work area.
In the “Visualizations” panel, select “Stacked column chart.”
Resize the chart by drabbing their move-handles.
Make sure the “Axis” box contains “order_dow” and the “Values” box with “order_id” respectively. Power BI Desktop should automatically calculate the count for “order_id” and display the field as “Count of order_id” as shown below.
The graph above is interesting because it shows a higher number of orders for Day 0 and Day 1.
Let’s make another chart.
We will follow the same procedure of adding a chart and for this time, we’ll use “order_hour_of_day” in the “Axis” box as shown below.
The graph shows the peak time for the number of orders.
One last graph!
We will add another chart with “days_since_prior_order” in the “Axis” box.
This last graph is the most interesting because the number of reorders peaks during these three time periods: 7 days, 14 days, and 30 days since prior order. This means that people are in a habit of resupplying every week, every two weeks, and every month.
That’s it, folks!
In the next article, we will “prettify” our charts and make them more readable to the others.
The procedures above have been drawn out. But if you’re a novice Power BI user, don’t despair! With regular practice, the concepts demonstrated in this article will soon become second nature and you’ll probably be able to do them in your sleep.
Let’s load the data into Power BI. Open up Power BI and click on “Get data” on the welcome screen as shown below.
Next, you’ll be presented with another pane that asks what type of data we want to get. Select “Text/CSV” as shown below and click on “Connect.”
When the File Open window appears, navigate to where we saved the dataset and click on the “Open” button on the lower right-hand corner.
When a preview appears, just click on “Load.”
We’ll now see the main working area of Power BI. Head over to the “Visualizations” panel and look for “Line Chart.”
This is what the line chart icon looks like:
Next, a visual placeholder will appear. Drag the hot corner marking on the lower right-hand corner of the placeholder and drag it diagonally down and to the right corner of the main working area.
Next, head over the “Fields” panel.
With the line chart placeholder still selected, find the “Date” field and click on the square box to put a checkmark on it.
We’ll now see the “Date” field under Axis. Click on the down arrow on the right of the “Date” as shown below.
Select “Date” instead of the default, “Date Hierarchy.”
Then, let’s put a checkmark on the “Births” field.
We’ll now see a line graph like the one below. Head over the Visualization panel and under the list of icons, find the analytics icon as shown below.
Scroll down the panel and find the “Forecast” section. Click on the down arrow to expand it if necessary.
Next, click on “+Add” to add forecasting on the current visualization.
We’ll now see a solid gray fill area and a line plot to the right of the visualization like the one below.
Let’s change the Forecast length to 31 points. In this case, a data point equals a day so 31 would roughly equate to a month’s worth of predictions. Click on “Apply” on the lower right-corner of the Forecast group to apply the changes.
Instead of points, let’s change the unit of measure to “Months” instead as shown below.
Once we click “Apply,” we’ll see the changes in the visualization. The graph below contains forecast for 3 months.
What if we wanted to compare how the forecast compares to actual data? We can do this with the “Ignore last” setting.
For this example, let’s ignore the last 3 months of the data. Power Bi will then forecast 3 months worth of data using the dataset but ignoring the last 3 months. This way, we can compare the Power BI’s forecasting result with the actual data in the last 3 months of the dataset.
Let’s click on “Apply” when we’redone changing the settings as shown below.
Below, we can see how the Power BI forecasting compares with the actual data. The black solid line represents the forecasting while the blue line represents the actual data.
The solid gray fill on the forecasting represents the confidence interval. The higher its value, the large the area will be. Let’s lower our confidence interval to 75% as shown below and see how it affects the graph.
The solid gray fill became smaller as shown below.
Next, let’s take into account seasonality. Below, let’s set it to 90 points which is equivalent to about 3 months. Putting this value will tell Power BI to look for seasonality within a 3-month cycle. Play with this value with what makes sense according to the data.
The result is show below.
Let’s return our confidence interval to the default value of 95% and scroll down the group to see formatting options.
Let’s change the forecasting line to an orange color and let’s make the gray fill disappear by changing formatting to “None.”
And that’s it! With a few simple clicks of the mouse, we got ourselves a forecast from the dataset.
Every once in a while, I would come across an article that decries online data science courses and boot camps as pathways towards getting a data science job. Most of the articles aim not to discourage but serve as a reminder to take a hard look in the mirror first and realize what we’re up against. However, a few detractors have proclaimed that the proliferation of these online courses and boot camps have caused the degradation of the profession.
To the latter, I vehemently disagree.
Bridging the Skill Gap
Data science have captured popular imagination ever since Harvard Business Review dubbed data scientist as the sexiest job of the 21st century. More than seven years later, data science remains one of the most highly sought-after job markets today. In fact, due to the dynamics of supply and demand, “the United States alone is projected to face a shortfall of some 250,000 data scientists by 2024¹.”
As a result, capitalism and entrepreneurship answered the call and companies like Codeup have vowed to “help bridge the gap between companies and people wanting to enter the field.”²
In addition, AutoML libraries like PyCaret are “democratizing machine learning and the use of advanced analytics by providing free, open-source, and low-code machine learning solution for business analysts, domain experts, citizen data scientists, and experienced data scientists”³.
The availability of online courses, boot camps, and AutoML libraries has led a lot of data scientists to raise their brows. They fear that boot camp alumni and self-taught candidates would somehow lower the overall caliber of data scientists and disgrace the field. Furthermore, they are afraid that the availability of tools like AutoML would allow anyone to be a data scientist.
I mean, God forbid if anyone thinks that they too can be data scientists! Right?
The Street Smart Data Scientist
Alumni of boot camps and self-taught learners, like myself, have one thing going for them: our rookie smarts. To quote Liz Wiseman, author of the book Rookie Smarts:
In a rapidly changing world, experience can be a curse. Being new, naïve, and even clueless can be an asset. — Liz Wiseman
Rookies are unencumbered. We are alert and constantly seeking like hunter-gatherers, cautious but quick like firewalkers, and hungry and relentless like frontiersmen⁴. In other words, we’re street smart.
Many are so bogged down by “you’ve got to learn this” and “you’ve got learn that” that they forget to stress the fact that data science is so vast that you can’t possibly know everything about anything. And that’s okay.
We learn fast and adapt quickly.
At the end of the day, it’s all about the value that we bring to our organizations. They are, after all, the ones paying our bills. We don’t get paid to memorize formulas or by knowing how to code an algorithm from scratch.
We get paid to solve problems.
And this is where the street smart data scientist excels. We don’t suffer from analysis paralysis or be bothered with theories, at least not while on the clock. Our center of focus is based on pragmatic solutions to problems, not on academic debate.
This is not to say we’re not interested in the latest research. In fact, it’s quite the contrary. We are voracious consumers of the latest development in machine learning and AI. We drool over the latest development in natural language processing. And we’re always on the lookout for the latest tool that will make our jobs easier and less boring.
So what if we have to use AutoML? If it gets us to an automatic pipeline where analysts can get the results of machine learning without manual intervention by a data scientist, the better. We’re not threatened by automation, we’re exhilarated by it!
Do not let perfection be the enemy of progress. — Winston Churchill
By building an automatic pipeline, there’s bound to be some tradeoffs. But building it this way will free our brain cells and gives us more time to focus on solving other higher-level problems and produce more impactful solutions.
We’re not concerned about job security, because we know that it doesn’t exist. What we do know is that the more value we bring to a business, the better we will be in the long run.
Maybe They’re Right?
After all this, I will concede a bit. For the sake of argument, maybe they’re right. Maybe online courses, boot camps, and low-code machine learning libraries really do produce low-caliber data scientists.
But still, I argue, this doesn’t mean we don’t have value. Data science skills lie on a spectrum and so does companies’ maturity when it comes to data. Why hire a six-figure employee when your organization barely has a recognizable machine learning infrastructure?
Maybe, to be labeled as a data scientist, one must be a unicorn first. A unicorn data scientist is a data scientist who excels at all facets of data science.
Data science has long been described as the intersection between computer science, applied statistics, and business or domain knowledge. To this, they ask, how can one person possibly accumulate all those knowledge into just a few months? To this, we also ask the same question, how can a college grad?
Unicorns do exist I believe, but they also have had to start from somewhere.
So why can’t we?
A whole slew of online courses and tools promise to democratize data science, and this is a good thing.
I was bored over the weekend so I decided to restore my Macbook Pro to factory settings so that I can set up my programming environment the proper way. After all, what’s a data scientist without her toys?
Let’s start with a replacement to the default terminal and pyenv installation to manage different Python versions.
Let’s move on to managing different Python interpreters and virtual environments using pyenv-virtualenv.
I remember a brief conversation with my boss’ boss a while back. He said that he wouldn’t be impressed if somebody in the company built a face recognition tool from scratch because, and I quote, “Guess what? There’s an API for that.” He then goes on about the futility of doing something that’s already been done instead of just using it.
This gave me an insight into how an executive thinks. Not that they don’t care about the coolness factor of a project, but at the end of that day, they’re most concerned about how a project will add value to the business and even more importantly, how quickly it can be done.
In the real world, the time it takes to build prototype matters. And the quicker we get from data to insights, the better off we will be. These help us stay agile.
And this brings me to PyCaret.
PyCaret is an open source, low-code machine learning library in Python that allows you to go from preparing your data to deploying your model within seconds in your choice of notebook environment.
Pycaret is basically a wrapper for some of the most popular machine learning libraries and frameworks scikit-learn and spaCy. Here are the things that PyCaret can do:
Natural Language Processing
Associate Rule Mining
If you’re interested in reading about the difference between traditional NLP approach vs. PyCaret’s NLP module, check out Prateek Baghel’s article.
Natural Language Processing
In just a few lines of code, PyCaret makes natural language processing so easy that it’s almost criminal. Like most of its other modules, PyCaret’s NLP module streamlined pipeline cuts the time from data to insights in more than half the time.
For example, with only one line, it performs text processing automatically, with the ability to customize stop words. Add another line or two, and you got yourself a language model. With yet another line, it gives you a properly formatted plotly graph. And finally, adding another line gives you the option to evaluate the model. You can even tune the model with, guess what, one line of code!
Instead of just telling you all about the wonderful features of PyCaret, maybe it’s be better if we do a little show and tell instead.
For this post, we’ll create an NLP pipeline that involves the following 6 glorious steps:
Getting the Data
Setting up the Environment
Creating the Model
Assigning the Model
Plotting the Model
Evaluating the Model
We will be going through an end-to-end demonstration of this pipeline with a brief explanation of the functions involved and their parameters.
Let’s get started.
Let us begin by installing PyCaret. If this is your first time installing it, just type the following into your terminal:
pip install pycaret
However, if you have a previously installed version of PyCaret, you can upgrade using the following command:
pip install —-upgrade pycaret
Beware: PyCaret is a big library so it’s going to take a few minutes to download and install.
We’ll also need to download the English language model because it is not included in the PyCaret installation.
Next, let’s fire up a Jupyter notebook and import PyCaret’s NLP module:
#import nlp module
from pycaret.nlp import *
Importing the pycaret.nlp automatically sets up your environment to perform NLP tasks only.
Getting the Data
Before setup, we need to decide first how we’re going to ingest data. There are two methods of getting the data into the pipeline. One is by using a Pandas dataframe and another is by using a simple list of textual data.
Passing a DataFrame
#import pandas if we're gonna use a dataframe
import pandas as pd
# load the data into a dataframe
df = pd.read_csv('hilaryclinton.csv')
Above, we’re simply loading the data into a dataframe.
Passing a List
# read a file containing a list of text data and assign it to 'lines'
with open('list.txt') as f:
lines = f.read().splitlines()
Above, we’re opening the file 'list.txt' and reading it. We assign the resulting list into the lines.
From the rest of this experiment, we’ll just use a dataframe to pass textual data to thesetup() function of the NLP module. And for the sake of expediency, we’ll sample the dataframe to only select a thousand tweets.
# sampling the data to select only 1000 tweets
df = df.sample(1000, random_state=493).reset_index(drop=True)
Let’s take a quick look at our dataframe with df.head() and df.shape.
Setting Up the Environment
In the line below, we’ll initialize the setup by calling the setup() function and assign it to nlp.
With data and target, we’re telling PyCaret that we’d like to use the values in the 'text' column of df. Also, we’re setting the session_id to an arbitrary number of 493 so that we can reproduce the experiment over and over again and get the same result. Finally, we added custom_stopwords so that PyCaret will exclude the specified list of words in the analysis.
Note that if we want to use a list instead, we could replace df with lines and get rid of target = ‘text’ because a list has no columns for the PyCaret to target!
Here’s the output of nlp:
The output table above confirms our session id, number of documents (rows or records), and vocabulary size. It also shows up if we used custom stopwords or not.
Creating the Model
Below, we’ll create the model by calling the create_model() function and assign it to lda. The function already knows to use the dataset that we specified during setup(). In our case, the PyCaret knows we want to create a model based on the 'text' in df.
# create the model
lda = create_model('lda', num_topics = 6, multi_core = True)
In the line above, notice that w param used 'lda' as the parameter. LDA stands for Latent Dirichlet Allocation. We could’ve just as easily opted for other types of models.
Here’s the list of models that PyCaret currently supports:
‘lda’: Latent Dirichlet Allocation
‘lsi’: Latent Semantic Indexing
‘hdp’: Hierarchical Dirichlet Process
‘rp’: Random Projections
‘nmf’: Non-Negative Matrix Factorization
I encourage you to research the difference between the models above, To start, check out Lettier’s awesome guide on LDA.
The next parameter we used is num_topics = 6. This tells PyCaret to use six topics in the results ranging from 0 to 5. If num_topic is not set, the default number is 4. Lastly, we set multi_core to tell PyCaret to use all available CPUs for parallel processing. This saves a lot of computational time.
Assigning the Model
By calling assign_model(), we’re going to label our data so that we’ll get a dataframe (based on our original dataframe: df) with additional columns that include the following information:
Topic percent value for each topic
The dominant topic
The percent value of the dominant topic
# label the data using trained model
df_lda = assign_model(lda)
Let’s take a look at df_lda.
Plotting the Model
Calling the plot_model() function will give us some visualization about frequency, distribution, polarity, et cetera. The plot_model() function takes three parameters: model, plot, and topic_num. The model instructs PyCaret what model to use and must be preceded by a create_model() function. topic_num designates which topic number (from 0 to 5) will the visualization be based on.
PyCarets offers a variety of plots. The type of graph generated will depend on the plot parameter. Here is the list of currently available visualizations:
‘frequency’: Word Token Frequency (default)
‘distribution’: Word Distribution Plot
‘bigram’: Bigram Frequency Plot
‘trigram’: Trigram Frequency Plot
‘sentiment’: Sentiment Polarity Plot
‘pos’: Part of Speech Frequency
‘tsne’: t-SNE (3d) Dimension Plot
‘topic_model’ : Topic Model (pyLDAvis)
‘topic_distribution’ : Topic Infer Distribution
‘wordcloud’: Word cloud
‘umap’: UMAP Dimensionality Plot
Evaluating the Model
Evaluating the models involves calling the evaluate_model() function. It takes only one parameter: the model to be used. In our case, the model is stored is lda that was created with the create_model() function in an earlier step.
The function returns a visual user interface for plotting.
And voilà, we’re done!
Using PyCaret’s NLP module, we were able to quickly from getting the data to evaluating the model in just a few lines of code. We covered the functions involved in each step and examined the parameters of those functions.
Thank you for reading! PyCaret’s NLP module has a lot more features and I encourage you to read their documentation to further familiarize yourself with the module and maybe even the whole library!
In the next post, I’ll continue to explore PyCaret’s functionalities.
If you want to learn more about my journey from slacker to data scientist, check out the article here.
I have a recurring dream where my instructor from a coding boot camp would constantly beat my head with a ruler telling me to read a package or library’s documentation. Hence, as a past time, I would find myself digging into Python or Panda’s documentation.
Today, I found myself wandering into pandas’ .drop() function. So, in this post, I shall attempt to make sense of panda’s documentation for the ever famous .drop().
Let’s import pandas and create a sample dataframe.
If we type df into a cell in Jupyter notebook, this will give us the whole dataframe:
One-level DataFrame Operations
Now let’s get rid of some columns.
df.drop(['color', 'score'], axis=1)
The code above simply tells Python to get rid of the 'color' and 'score' in axis=1 which means look in the columns. Alternatively, we could’ve just as easily not used the named parameter axis because it’s confusing. So, let’s try that now:
Both of the methods above will result in the following:
Next, we’ll get rid of some rows (or records).
df.drop([1, 2, 4, 6])
Above, we’re simply telling Python to get rid of the rows with the index of 1, 2, 4, and 6. Note that the indices are passed as a list [1, 2, 4, 6]. This will result in the following:
MultiIndex DataFrame Operations
In this next round, we’re going to work with a multi-index dataframe. Let’s set it up:
Next, let’s get rid of 'pork rinds' because I don’t like them:
df.drop(index='pork rinds', level=0)
And this is what we get:
And finally, let’s cut the fat:
Above, level=1 simply means the second level (since the first level starts with 0). In this case, it’s the carbs, fat, and protein levels. By specifying index='fat', we’re telling Python to get rid of the fat in level=1.
Here’s what we get:
So far, with all the playing that we did, somehow, if we type df into a cell, the output that we’re going to get is the original dataframe without modifications. this is because all the changes that we’ve been making take effect only on the display.
But what if we want to make the changes permanent? Enter: inplace.
df.drop(index='fat', level=1, inplace=True)
Above, we added inplace=True in the parameter. This signals Python that we want the changes to be taken in place so that when we output df, this is what we’ll get: