networkx-saas / app.py
LeonceNsh's picture
Update app.py
a5ad05e verified
raw
history blame
8.92 kB
import pandas as pd
import networkx as nx
import matplotlib.pyplot as plt
from io import BytesIO
from PIL import Image
import gradio as gr
import logging
# Set up logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
# Load and preprocess the dataset
file_path = "cbinsights_data.csv" # Replace with your actual file path
try:
data = pd.read_csv(file_path, skiprows=1)
logger.info("CSV file loaded successfully.")
except FileNotFoundError:
logger.error(f"File not found: {file_path}")
raise
except Exception as e:
logger.error(f"Error loading CSV file: {e}")
raise
# Standardize column names: strip whitespace and convert to lowercase
data.columns = data.columns.str.strip().str.lower()
logger.info(f"Standardized Column Names: {data.columns.tolist()}")
# Identify the valuation column dynamically
valuation_columns = [col for col in data.columns if 'valuation' in col.lower()]
if not valuation_columns:
logger.error("No column containing 'Valuation' found in the dataset.")
raise ValueError("Data Error: Unable to find the valuation column. Please check your CSV file.")
elif len(valuation_columns) > 1:
logger.error("Multiple columns containing 'Valuation' found in the dataset.")
raise ValueError("Data Error: Multiple valuation columns detected. Please ensure only one valuation column exists.")
else:
valuation_column = valuation_columns[0]
logger.info(f"Using valuation column: {valuation_column}")
# Clean and prepare data
data["valuation_billions"] = data[valuation_column].replace({'\$': '', ',': ''}, regex=True)
data["valuation_billions"] = pd.to_numeric(data["valuation_billions"], errors='coerce')
logger.info("Valuation data cleaned and converted to numeric.")
# Strip whitespace from all string columns
data = data.apply(lambda col: col.str.strip() if col.dtype == "object" else col)
logger.info("Whitespace stripped from all string columns.")
# Rename columns for consistency
expected_columns = {
"company": "Company",
"valuation_billions": "Valuation_Billions",
"date_joined": "Date_Joined",
"country": "Country",
"city": "City",
"industry": "Industry",
"select_investors": "Select_Investors"
}
missing_columns = set(expected_columns.keys()) - set(data.columns)
if missing_columns:
logger.error(f"Missing columns in the dataset: {missing_columns}")
raise ValueError(f"Data Error: Missing columns {missing_columns} in the dataset.")
data = data.rename(columns=expected_columns)
logger.info("Columns renamed for consistency.")
# Parse the "Select_Investors" column to map investors to companies
def build_investor_company_mapping(df):
mapping = {}
for _, row in df.iterrows():
company = row["Company"]
investors = row["Select_Investors"]
if pd.notnull(investors):
for investor in investors.split(","):
investor = investor.strip()
if investor: # Ensure investor is not an empty string
mapping.setdefault(investor, []).append(company)
return mapping
investor_company_mapping = build_investor_company_mapping(data)
logger.info("Investor to company mapping created.")
# Function to filter investors based on selected country and industry
def filter_investors_by_country_and_industry(selected_country, selected_industry):
filtered_data = data.copy()
logger.info(f"Filtering data for Country: {selected_country}, Industry: {selected_industry}")
if selected_country != "All":
filtered_data = filtered_data[filtered_data["Country"] == selected_country]
logger.info(f"Data filtered by country: {selected_country}. Remaining records: {len(filtered_data)}")
if selected_industry != "All":
filtered_data = filtered_data[filtered_data["Industry"] == selected_industry]
logger.info(f"Data filtered by industry: {selected_industry}. Remaining records: {len(filtered_data)}")
investor_company_mapping_filtered = build_investor_company_mapping(filtered_data)
# Calculate total valuation per investor
investor_valuations = {}
for investor, companies in investor_company_mapping_filtered.items():
total_valuation = filtered_data[filtered_data["Company"].isin(companies)]["Valuation_Billions"].sum()
if total_valuation >= 20: # Investors with >= 20B total valuation
investor_valuations[investor] = total_valuation
logger.info(f"Filtered investors with total valuation >= 20B: {len(investor_valuations)}")
return list(investor_valuations.keys()), filtered_data
# Function to generate the graph
def generate_graph(selected_investors, filtered_data):
if not selected_investors:
logger.warning("No investors selected. Returning None for graph.")
return None
investor_company_mapping_filtered = build_investor_company_mapping(filtered_data)
filtered_mapping = {inv: investor_company_mapping_filtered[inv] for inv in selected_investors if inv in investor_company_mapping_filtered}
logger.info(f"Generating graph for {len(filtered_mapping)} investors.")
# Build the graph
G = nx.Graph()
for investor, companies in filtered_mapping.items():
for company in companies:
G.add_edge(investor, company)
# Node size based on valuation
max_valuation = filtered_data["Valuation_Billions"].max()
node_sizes = []
for node in G.nodes:
if node in filtered_mapping:
node_sizes.append(1500) # Fixed size for investors
else:
valuation = filtered_data.loc[filtered_data["Company"] == node, "Valuation_Billions"].sum()
size = (valuation / max_valuation) * 1500 if max_valuation else 100
node_sizes.append(size)
# Node color: Investors (orange), Companies (green)
node_colors = ["#FF8C00" if node in filtered_mapping else "#32CD32" for node in G.nodes]
# Draw the graph
plt.figure(figsize=(15, 15))
pos = nx.spring_layout(G, k=0.2, seed=42)
nx.draw(
G, pos,
with_labels=True,
node_size=node_sizes,
node_color=node_colors,
font_size=10,
edge_color="#A9A9A9", # Light gray edges
alpha=0.9
)
# Legend
from matplotlib.lines import Line2D
legend_elements = [
Line2D([0], [0], marker='o', color='w', label='Investor', markersize=10, markerfacecolor='#FF8C00'),
Line2D([0], [0], marker='o', color='w', label='Company', markersize=10, markerfacecolor='#32CD32')
]
plt.legend(handles=legend_elements, loc='upper left')
plt.title("Venture Network Visualization", fontsize=20)
plt.axis("off")
# Save plot to BytesIO
buf = BytesIO()
plt.savefig(buf, format="png", bbox_inches="tight")
plt.close()
buf.seek(0)
logger.info("Graph generated successfully.")
return Image.open(buf)
# Gradio app function
def app(selected_country, selected_industry):
investor_list, filtered_data = filter_investors_by_country_and_industry(selected_country, selected_industry)
logger.info("Updating CheckboxGroup and filtered data holder.")
# Use gr.update() to create an update dictionary for the CheckboxGroup
return gr.update(
choices=investor_list,
value=investor_list,
visible=True
), filtered_data
# Gradio Interface
def main():
country_list = ["All"] + sorted(data["Country"].dropna().unique())
industry_list = ["All"] + sorted(data["Industry"].dropna().unique())
logger.info(f"Available countries: {country_list}")
logger.info(f"Available industries: {industry_list}")
with gr.Blocks() as demo:
with gr.Row():
# Set default value to "US" for country and "Enterprise Tech" for industry
country_filter = gr.Dropdown(choices=country_list, label="Filter by Country", value="United States")
industry_filter = gr.Dropdown(choices=industry_list, label="Filter by Industry", value="Enterprise Tech")
filtered_investor_list = gr.CheckboxGroup(choices=[], label="Select Investors", visible=False)
graph_output = gr.Image(type="pil", label="Venture Network Graph")
filtered_data_holder = gr.State()
country_filter.change(
app,
inputs=[country_filter, industry_filter],
outputs=[filtered_investor_list, filtered_data_holder]
)
industry_filter.change(
app,
inputs=[country_filter, industry_filter],
outputs=[filtered_investor_list, filtered_data_holder]
)
filtered_investor_list.change(
generate_graph,
inputs=[filtered_investor_list, filtered_data_holder],
outputs=graph_output
)
demo.launch()
if __name__ == "__main__":
main()