Achievement π :
1st Place Overall & $7,000 USD
This competition was organized by Anaconda, the world's most popular data science platform. It was held in Singapore, at the National University of Singapore (NUS), where both Anaconda from the United States and AI Singapore jointly carried out 2 rounds of judging.
The exploration of the happiness topic stems from a profound curiosity to decode the intricate factors that contribute to our overall well-being and contentment. In our quest to understand this complex emotion and universal pursuit, we analyzed 5 datasets and visualized regional-level insights about happiness by coding 8 static and interactive charts using Python. We then weaved our charts and insights into a visual narrative (see poster) which tells its data-driven story about happiness! π»
Refer to the Datasets folder to download the following datasets:
- social_progress_dataset
- DataForTable2.1WHR2023
- undesa_destination_and_origin_processed
- hapiness_report_2022
- Regions
To install the required Python libraries/packages in your Jupyter Notebook, run:
!pip install pandas matplotlib seaborn scipy numpy plotly kaleido geopandas
This will install the following libraries along with their dependencies:
Pandas: An open-source data analysis and manipulation tool
Matplotlib: A comprehensive library for creating static, animated, and interactive visualizations in Python
Seaborn: A statistical data visualization library built on top of matplotlib
Scipy: An open-source Python library used for scientific and technical computing
Numpy: The fundamental package for scientific computing with Python
Plotly: An interactive graphing library for Python
Kaleido: For static image export support with Plotly
Geopandas: An open-source project that makes working with geospatial data in python easier
- Polar Histogram
- Choropleth Map
- Histogram with KDE
- Radar Chart
- Boxplot
- Sankey Diagram
- Bar Chart
- Scatterplot
The go.Sankey function consists of 2 parameters node (dict) and link (dict). The primary technical challenge in generating the Sankey diagram was constructing the inputs for sub-parameters source, target and value in the link (dict) parameter. And the inputs required an advanced DataFrame that the function could interpret.
matrix_inc = df_inc.iloc[:, 1:].values.tolist()
names_inc = df_inc.columns[1:].tolist()
data_inc = [{'source': names_inc.index(name), 'target': names_inc.index(target_name), 'value': matrix_inc[i][j]}
for i, name in enumerate(names_inc)
for j, target_name in enumerate(names_inc) if matrix_inc[i][j] > 0]
We achieved the desired format by using a list comprehension to iterate over our extracted matrix data, transforming it into a list data_inc where each element is a dictionary representing a link between two nodes (countries). The comprehension for i, name in enumerate(names_inc) iterates over each country name to establish it as the source node, while a nested loop for j, target_name in enumerate(names_inc) determines each target node. The key 'source' is set to names_inc.index(name), leveraging the index() function to translate the country name to its corresponding index in the names_inc list.
Likewise, the 'target' key is assigned the index of the target country, and the 'value' key holds the migration flow from the source to the target extracted from matrix_inc[i][j]. In short, this indexing and mapping method converts country names into numerical indices that are used to draw the links between nodes, thereby representing the flow of migrants between different income-level countries.
In developing the polar histogram, we encountered with challenges with positioning the country names and scores as such: <score> <country> for the right hemisphere and <country> <score> for the left hemisphere of the plot, while also aligning and rotating with the positions of the data points.
for theta_val, (country, score) in zip(theta_with_10_degree_gap, data_sorted[['country', 'score']].values):
alignment = {}
if 0 <= theta_val <= np.pi/2 or 3*np.pi/2 <= theta_val <= 2*np.pi:
text_content = f" ({score:.2f}) {country}"
else:
text_content = f"{country} ({score:.2f}) "
if np.pi/2 < theta_val < 3*np.pi/2:
alignment = {"ha": "right", "va": "center"}
rotation_deg = theta_val*(180/np.pi) - 180
else:
alignment = {"ha": "left", "va": "center"}
rotation_deg = theta_val*(180/np.pi)
The code first determines the text content's sequence text_content based on the angular position theta_val of the data point. For data points in the right hemisphere (0 to Ο/2 and 3Ο/2 to 2Ο radians), the text is formatted as <score> <country>. For points in the left hemisphere (Ο/2 to 3Ο/2 radians), it's <country> <score>. Depending on whether the data point is in the left or right hemisphere, the horizontal alignment ha is set to "right" or "left", respectively. This ensures that the text is aligned in a way that it's always outward-facing from the center of the plot, enhancing readability (vertical alignment va is consistently set to "center").
The rotation of the text rotation_deg is calculated based on theta_val, with an adjustment of - 180 degrees when the point is in the left hemisphere. This rotation aligns the text with the radial lines of the plot, ensuring that the text orientation is consistent with the direction of the data points.
Thank you for reading! π