As a wine lover, I wanted to explore wine data from Kaggle to provide consumers the ability to make data driven decisions when purchasing wine. Data set link: https://www.kaggle.com/zynicide/wine-reviews
The data set I chose is composed of 130,000 wine twitter reviews. The dataset provides the name of the wine, points the wine receives, price of the wine, country, region, and province of the wines. I took a random subset of 15,000 wine reviews from the Kaggle data set to then perform my analysis.
According to https://www.winespectator.com/display/show/id/scoring-scale, wine ratings are not based on price and rated as:
50-59 wines are undrinkable and not recommended,
60-69 wines are flawed and not recommended but drinkable,
70-79 wines are flawed and taste average,
80-84 wines are ‘above average’ to ‘good’,
85-90 wines are ‘good’ to ‘very good’,
90-94 wines are ‘superior’ to ‘exceptional’
95-100 wines are benchmark examples or ‘classic’.
When walking into the wine store, wines are normally categorized by the country of origin. I wanted to demonstrate with a bubble chart, average price versus average points of wines by country. The size of the bubbles represent the amount of wine reviews per country received.
I wanted to demonstrate the dataset is skewed towards American wines accounting for 45.5% of all the reviews , followed by French wines accounting for 14% of all wine reviews, then Italian wines reflecting 13.4% of of the wine reviews.
In the first chart you can see there is a correlation between price and points, the lower the points the cheaper the wine. In this data set the lowest rating reviewed was 80. The outliers are interesting and can allow the consumer to view other options of wine in the same rating range as the US, France, and Italy but on average less expensive. A consumer does not need to sacrifice points they can simply purchase wine from another country they may not traditionally purchase. Canadian wine anyone?
The data did not provide the year the wine was made, I had to extract that information from the title of the wine. The vintage year is a very important component about wine. The grape harvest is extremely dependent on climate and weather. Most wines are meant to be enjoyed right away or shelf life of 4-5 years, only 1% of wines are meant for consumption over 5 years.
I thought it would be interesting to see how wine prices change depending on the year the wine was made. The graph portrays average price per year. The older the wine the more expensive it is, it is costly to maintain older wines as they must avoid exposure to light, and stored in 65-70 degrees. In more recent years the data demonstrates that 2001, & 2016 on average were cheaper than the immediate years after. There are various reasons, winemakers changes their methods, climate affects the grapes greatly, grapes my not become ripe enough, too much rain dilutes the grape juice.
I wanted the viewer to see the a histogram on the distribution of points, most wine falls in this data set falls in the category of 88 to 90 points. There are not many wines that reach 100 points as a rating as this is extremely difficult to get.
Typically when a wine expert reviews wine they are trained to taste flavors the average consumers typically cannot detect. My word cloud demonstrates the words most frequently used to describe wines. The user can also filter by blend, country, and region. The user can view the words used to describe their favorite wines, in turn then search for other wines that may have similar flavors.
I lastly wanted to make this application practical in order for the user to find the wine they seek. The interactive data set allows users to search wine by country, province, region, variety, price range, and points range. The user is able to search by words used to describe the wines.
I hope you find your perfect wine!
Link to my shiny app: https://github.com/ariani86/WineShinyApp.git