Data Analysis of Food Network Recipes
The collected data contained approximately 4000 recipes with rating and comments where it was available. The data included name, cooking time, level of difficulty and ingredients. The cleaning included dealing with missing information, converting text to numerical values. The comment section contained web links, recipes and emojis, therefore I left it out of the analysis this time.
Out of many free time activities cooking is one of my favorite one. It is relaxing as it really turns off the wheels in my head. I focus on the meal fully with engineering detail orientation. Also, the result of the work is a lovely satisfying meal, which I like to share with friends or family.
That is why I chose collecting recipes and related data from my favorite channel's website foodnetwork.com for my project. I scraped the website with Scrapy and used Selenium for the dynamic content.
First, I checked the items with the higher rating. Interestingly, majority of the ratings were rather high. I wondered if all the recipes tastes really that good. Perhaps, those who did not find the recipe that good simply did not comment at all. See the overall rating chart in form of density curve.
Next, I reviewed the total preparation time of the recipes. What are is the most frequent preparation time for a recipe?
While converting the text field of hour and minute to numbers I had some troubles with my conversion function. I realized that there are recipes with days of preparation time. I had to check what takes so long to prepare and listed them in a table. Lemon-Lavender pie was the overall winner over 1600 minutes of preparation time.
I also looked into the serving sizes of recipes. I visualized them in a histogram chart.
Most serving sizes are between five to ten, yet again my chart has data points up to hundred servings. So, as I did before I wanted to find out what are those outliers. I listed them in a chart below.
I also wanted to find answer to a couple of questions. Does it take a long time to prepare meals with large serving size? From the scatter plot below, one is able to see that there is no correlation between preparation time and preparation of large servings.
Lastly, but not less importantly let us see if cooks like to prepare and taste complex meals. Do users rate easy meals high or more complex meals with perhaps more flavors. As it turns out, there is not apparent variation in ratings depending on the number of ingredients needed for the recipe. However, this tendency cannot be stated with a high confidence.
While, this project did not reveal shocking discoveries, it did give some insights into interesting trends that high ratings does not necessarily mean that the recipe is great, that it takes a long time to prepare complex meals or that it is possible to prepare meals for many souls in a short period of time.
It was a nice practice for a visualizing exercise and hoping to extend my analysis into comments content by applying NLP and how ingredients are relevant to ratings.
The skills the authors demonstrated here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
About Author
