For our project we use 3 datasets: (1) scraped data from the website Kofio, (2) Kaggle Coffee Dataset, (3) collected data about Coffee-Food-Pairing.
Kofio Dataset. We scraped the necessary data from Kofio because we couldn't find a suitable dataset that met our requirements regarding coffee flavors, brewing methods, and brands. Kofio offers a wide range of coffee products from various European brands. All of the coffee products on Kofio have the same set of information, as an example:
| Key | Value |
|---|---|
| Item Name: | EASTER FILTER blend - Easter limited edition |
| Roast Date: | 27.03.2023 (8 days) |
| Roastery: | Square Mile |
| Coffee Origin: | Kenya specialty coffee Kenya, Peru specialty coffee Peru |
| Region: | Kirinyaga |
| Variety: | Caturra |
| Roast Type: | Filter |
| Process: | Washed |
| Flavour Profile: | Apple blossom, Cherry pie |
| Roast Level: | Light to Medium Light |
| Brewing Method: | French Press, Hario V60, Vacuum Pot |
| Rating: | 4.4 |
| Recommended: | 93% |
| Price: | 19,99 |
Besides the above information listed in the form, each product also has a price, a rating, and a recommendation rate. We also crawled image for each product for later possible use. Although each coffee product has exactly the same set of keys, some of the values are missed, thus requiring missing value treatment. Besides, the Price variable should be of float type instead of string with comma, so we convert them to float. All of these(crawlers & data analysis) are done in Python.
Coffee dataset. We use the Coffee Dataset to display the global production and consumption of coffee. This dataset provides information on the amount of coffee export, import and consumption by countries from 1990 to 2019.
Pairing dataset. We collected the data about how to pair coffee with data from Homegrounds website. The dataset we collected has three columns: Coffee origin, Food and Flavor type. For example:
| Coffee origin | Food | Flavor type |
|---|---|---|
| Yemen | Blueberries | Fruity |
| Columbia | white chocolate | Chocolate |
The dataset is clean and needn't to be preprocessed.
Originating from the mysterious land of the Kingdom of Kaffa in Eutopia, coffee has evolved into much more than a tasty and healthy beverage. It has transcended into a cultural symbol, a social ritual, a source of comfort, and a muse for inspiration. Across the globe, coffee shops have transformed into meeting places for friends, colleagues, and neighbors, fostering a sense of community. Additionally, coffee production is a significant economic driver in numerous societies, offering income and employment opportunities to small-scale farmers and large-scale plantations.
Recognizing coffee's prestige, our project Kingdom of Kaffa provides a comprehensive overview of coffee's impact, including its various origins, flavors, and brewing techniques. Our website also serves as a platform for users to find their ideal and real European coffee brand, facilitating their immersion into the coffee community. Our website caters to coffee enthusiasts and beginners, allowing everyone to find their perfect cup of coffee.
To meet the goals describe above, our website will roughly (might change sligthly overtime) contain the following visualizations:
Figure 1: This map-based visualization showcases the production and consumption levels of coffee at the country level over an extended period of time. This visualization allows users to examine the fluctuations in demand for coffee, the global distribution of coffee plantations, and the countries with the highest consumption rates.
Figure 2: This graph depicts the correlation between price, rating, and recommendation rates among a selection of European coffee brands. Each brand is visually distinguishable by unique colors, and users are afforded the option to adjust the x and y axis, allowing them to compare data points and make informed decisions based on their preferences.
Figure 3: This bubble-cluster visualization offers filtering functionalities based on four distinct indicators, namely flavor, brewing method, roast type, and roast level (subject to revision). Users are able to input their preferences for each indicator separately, and the figure will subsequently generate a filtered product that best conforms to their individual taste.
Figure 4: This pairing visulaization shows which types of food are recommended to be consumed with specific types of coffee, based on the origin/flavour of the coffee. The figure is structured in three columns, each of which corresponds to 'coffee origin', 'food', and 'flavor type'. The graph is interactive, allowing users to click on any bubble to establish connections with other relevant bubbles, represented by lines of the same color. Non-activating lines will be displayed in gray. Additionally, detailed and well-designed contents related to each bubble will be available upon clicking, offering users relevant information.
In the section Conda enviroment for EDA you can find instructions to install conda environment to run our python notebooks with EDA.
We will be using the Kofio dataset we scraped to create visualizations (Figure 2,3,4) about the typical coffee products. Detailed analysis can be found in the jupyter_notebook kofio analysis.
Kofio dataset contains 322 coffee products from 35+ brands. Each coffee product can be classified by Roast Level, Roast type, Brewing method and Flavour. In total, we gathered the following stats about each category:
There're 194 unique flavours: ['Flower honey', 'Black tea with lemon', 'Coke']...
There're 8 unique brewing methods: ['Vacuum Pot', 'Hario V60', 'Clever dripper', 'Espresso', 'Chemex', 'Moka pot', 'Aeropress', 'French Press']
There're 3 unique Roast Levels: ['Light to Medium Light', 'Medium to medium dark', 'Omni']
There're 3 unique Roast Types: ['Filter', 'Espresso', 'Omni']
In the figure below we display coffee product (each color and size depicts different brand) with respect to price/rate/recommendation. You can see that the price of a coffee brand is not always an indicator of its quality or popularity, other factors should be considered such as ratings and recommendations. Example, "yellow" brand is cheap but with a high rating while "light violet" are more expensive and average.
We analyze the pairing dataset using jupyter notebook pairing_analysis.
The dataset contains 83 entries and includes information on 25 different coffee origins, 23 different foods, and 8 different flavor types. We group the data by coffee origins and flavor types to see the origin-flavor, origin-food, food-flavor pairings.
The figure below presents a clear look at the origin-flavor pairings.
We analyze the Coffee dataset using jupyter notebook coffee_analysis.
We use Coffee dataset to visualize a world map of coffee consumption and produstion (Figure 1). The dataset includes seven files about coffee import and export, but we focused on the following two files:
- Coffee_total_consumption.csv - combination of Coffee_domestic_consumption.csv and Coffee_import_consumption.csv
- Coffee_production.csv
We used the geopandas library to create the map, but had to rename some country names in our dataset to match the library's names.
Global production.
This figure displays global coffee production rates for Arabica and Robusta coffee. Arabica and Robusta coffee types are produced in countries across Central and South America, Africa, and Asia, with some countries producing both. Brazil is the largest producer of both types, while Colombia and Côte d'Ivoire are the largest producers of Robusta and Arabica, respectively.
Global consumption.
The figure below suggests that coffee consumption is becoming increasingly popular around the world, but that the Americas may be the region with the highest demand for coffee.
Since the Kofio Dataset is scraped by us, nobody has done with this dataset. However, several works have conducted data analysis and visualization on the Coffee dataset.
- Coffee Economic EDA tabulates the total coffee consumption for 55 countries and subsequently displays the resulting data in a histogram format. The histogram portrays the consumption of each country in descending order. Besides, it analyzes the mean, median, and total consumption from 1990 to 2020 and the coffee type of each country.
- Coffee - Extensive EDA examines different aspects of the Coffee dataset, such as global coffee production and consumption, the top coffee-producing countries, the different types of coffee produced, and the top coffee importers and exporters. The analysis also includes a comparison of coffee production and consumption trends over the years. The resultant data was subsequently represented using line and bar charts.
- Simple EDA, Data Visualization of the Coffee Dataset produces a similar analysis as the work above.
Our approach is original due to the following aspects:
- We present coffee consumption and import/export volume trends across regions and their temporal evolution. This enables the display of increasing coffee popularity by year and helps identify the regions with the highest affinity for coffee.
- We analyze the existing coffee brands and consider the rich coffee flavors.
- Our visualization can help visitors to buy an existing coffee they love using the filtering function.
- We show clearly how to pair coffee to food and to flavor.
Also, several great visualizations from other topics have inspired us to come up with our ideas.
- Covid-19 in Switzerland. The SWITZERLAND MAP visualization inspired us to come up with our Figure 1. We want to visualize the change of coffee consumption, import volume and export volumn according to years and countries in a way which shows the growing popularity of coffee.
- Wine101. The cluster visualization inspired us to show clusters based on flavor, brewing method, and roast level.
- Coffee Flavor Wheel. This coffee flavor wheel inspired us to show the coffee-flavor, coffee-food or flavor-food pairings.
- Update conda packages:
conda update --all
- Create a conda environment from the provided yml file. It might take a couple of minutes.
conda env create --file=com480-kaffa.yml
- Activate the conda environment in your Code Editor. In VS code, choose com480-kaffa kernel in the python notebook. If it does not appear in the list of kernels after step 2, restart VS.