A common business analytics task is trying to forecast the future based on known historical data. Forecasting is a complicated topic and relies on an analyst knowing the ins and outs of the domain as well as knowledge of relatively complex mathematical theories. Because the mathematical concepts can be complex, a lot of business forecasting approaches are “solved” with a little linear regression and “intuition.” More complex models would yield better results but are too difficult to implement.

Given that background, I was very interested to see that Facebook recently open sourced a python and R library called prophet which seeks to automate the forecasting process in a more sophisticated but easily tune-able model. In this article, I’ll introduce prophet and show how to use it to predict the volume of traffic in the next year for Practical Business Python. To make this a little more interesting, I will post the prediction through the end of March so we can take a look at how accurate the forecast is.

Read more...

In a previous post, I covered one approach for generating documents using HTML templates to create a PDF. While PDF is great, the world still relies on Microsoft Word for document creation. In reality, it will be much simpler for a business user to create the desired template that supports all the custom formatting they need in Word versus trying to use HTML+CSS. Fortunately, there is a a package that supports doing a MS Word mailmerge purely within python. This approach has the advantage of running on any system - even if Word is not installed. The benefit to using python for the merge (vs. an Excel sheet) is that you are not limited in how you retrieve or process the data. The full flexibility and power of the python ecosystem is at your finger tips. This should be a useful tool to keep in mind any time you need to automate document creation.

Read more...

In many practical Data Science activities, the data set will contain categorical variables. These variables are typically stored as text values which represent various traits. Some examples include color (“Red”, “Yellow”, “Blue”), size (“Small”, “Medium”, “Large”) or geographic designations (State or Country). Regardless of what the value is used for, the challenge is determining how to use this data in the analysis. Many machine learning algorithms can support categorical values without further manipulation but there are many more algorithms that do not. Therefore, the analyst is faced with the challenge of figuring out how to turn these text attributes into numerical values for further processing.

As with many other aspects of the Data Science world, there is no single answer on how to approach this problem. Each approach has trade-offs and has potential impact on the outcome of the analysis. Fortunately, the python tools of pandas and scikit-learn provide several approaches that can be applied to transform the categorical data into suitable numeric values. This article will be a survey of some of the various common (and a few more complex) approaches in the hope that it will help others apply these techniques to their real world problems.

Read more...

Several months ago, I participated in my first crowd-sourced Data Science competition in the Twin Cities run by Analyze This!. In my previous post, I described the benefits of working through the competition and how much I enjoyed the process. I just completed the second challenge and had another great experience that I wanted to share and (hopefully) encourage others to try these types of practical challenges to build their Data Science/Analytics skills.

In this second challenge, I felt much more comfortable with the actual process of cleaning the data, exploring it and building and testing models. I found that the python tools continue to serve me well. However, I also identified a lot of things that I need to do better in future challenges or projects in order to be more systematic about my process. I am curious if the broader community has tips or tricks they can share related to some of the items I will cover below. I will also highlight a few of the useful python tools I used throughout the process. This post does not include any code but is focused more on the process and python tools for Data Science.

Read more...

In my last article, I discussed building a financial model in pandas that could be used for multiple amortization scenarios. Unfortunately, I realized that I made a mistake in that approach so I had to rethink how to solve the problem. Thanks to the help of several individuals, I have a new solution that resolves the issues and produces the correct results.

In addition to posting the updated solution, I have taken this article as an opportunity to take a step back and examine what I should have done differently in approaching the original problem. While it is never fun to make a mistake in front of thousands of people, I’ll try to swallow my pride and learn from it.

Read more...