Data Analysis | How to Win Over the Blind Date
The skills I demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
In the era of online dating, I still appreciate the romance brought by a blind date. Two people who have not previously met each other, and they travel from different parts of the city to the same place.
That is the reason why I like reading the column"The Undateables" on Timeout New york. "The Undateables" picks up two strangers to go on a blind date together. After a romantic dinner (or not), some awkward or hit-off moments, each couple gives a rating to the blind date.
I began wondering: What makes a date well, and what makes it go wrong? These are interesting questions to think.
Therefore, I started my analysis to figure the answers out.
Data Collection
I scraped the following information on each page of "The Undateables": The professions and living neighborhood of the couple, their ratings for each other, the restaurant they dated and its zip code.
The data I collected is from 2018 to 2019.
Feature Engineering
1) Professions and Wage: I searched the average wage in NYC for each profession on glassdoor, and create a column of "wage" based on the information.
2) The neighborhood, Distance, and Minute Commute: Based on the neighborhood, I have encoded them with their distance to the center of Newyork (Grand Central Station). Moreover, with the help of google map, I have calculated the minute's commute by subway between where they live.
3) Wage/Distance: This is one variable I create for trying to see the correlation between these two variables.
General Analysis
In the left graph, the right to the left means the data from 2018 to 2019. In the right graph, the bar chart means the proportion of high ratings (ratings range from 1-5, high ratings mean 4-5.)
Wage/Distance Analysis
The X-axis means wage/distance/1000. The Y-axis means rating from 1-5.
The rectangular in the left means high ratings concentrate in this part. Also, the perfect range is from 0-0.82. When wage/distance/1000 is from 0-0.82, that means the couple can have more possibilities in high ratings.
The three red points in the circle are outlined as outliers are also intriguing. Even if their scores for wage and distance are not perfect, they still have high ratings.
The three cases are:
couple are from different wage levels and unrelated industries, the dating can still be successful because of:
1. On-time or not mad at someone's being late
"She arrived later. That's not a problem for me."
"We got there at the same time and sat at the bar."
2. Being cute but also reserved is the right attitude.
"Physically, he was my type, but he seemed a bit reserved at first."
"But this was like a slow, building chemistry."
3. Talk is much better than not. Have as many good conversations as possible.
"I think I was kind of rambling, but he was actually really good at steering the conversation. "
Geography
I have also made one map for showing the distribution of restaurant where blind dates with high rating happened.
And here are the top restaurants where high rating dates happened.
This article is the first episode about the data analysis on a blind date, and I will post the second one soon on building a machine learning model for how to make the successful blind date.
Hope this article can be helpful for single birds in New York City. Happy Dating!