Update app.py

app.py

@@ -22,3 +22,154 @@ from datasets import load_dataset
 
 geo = load_dataset('jamescalam/world-cities-geo', split='train')
 st.write(geo)
+
+
+
+
+
+
+import plotly.express as px
+
+palette = ['#1c17ff', '#faff00', '#8cf1ff', '#000000', '#030080', '#738fab']
+
+fig = px.scatter_3d(
+    x=geo['x'], y=geo['y'], z=geo['z'],
+    color=geo['continent'],
+    custom_data=[geo['country'], geo['city']],
+    color_discrete_sequence=palette
+)
+fig.update_traces(
+    hovertemplate="\n".join([
+        "city: %{customdata[1]}",
+        "country: %{customdata[0]}"
+    ])
+)
+
+
+fig.write_html("umap-earth-3d.html", include_plotlyjs="cdn", full_html=False)
+
+
+import numpy as np
+
+geo_arr = np.asarray([geo['x'], geo['y'], geo['z']]).T
+geo_arr = geo_arr / geo_arr.max()
+
+st.markdown(geo_arr[:5])
+
+
+import umap
+
+
+colors = geo['continent']
+c_map = {
+    'Africa': '#8cf1ff',
+    'Asia': '#1c17ff',
+    'Europe': '#faff00',
+    'North America': '#738fab',
+    'Oceania': '#030080',
+    'South America': '#000000'
+}
+for i in range(len(colors)):
+    colors[i] = c_map[colors[i]]
+colors[:5]
+
+
+
+import matplotlib.pyplot as plt
+import seaborn as sns
+from tqdm.auto import tqdm
+
+fig, ax = plt.subplots(3, 3, figsize=(14, 14))
+nns = [2, 3, 4, 5, 10, 15, 30, 50, 100]
+i, j = 0, 0
+for n_neighbors in tqdm(nns):
+    fit = umap.UMAP(n_neighbors=n_neighbors)
+    u = fit.fit_transform(geo_arr)
+    sns.scatterplot(x=u[:,0], y=u[:,1], c=colors, ax=ax[j, i])
+    ax[j, i].set_title(f'n={n_neighbors}')
+    if i < 2: i += 1
+    else: i = 0; j += 1
+
+
+target = geo['continent']
+t_map = {
+    'Africa': 0,
+    'Asia': 1,
+    'Europe': 2,
+    'North America': 3,
+    'Oceania': 4,
+    'South America': 5
+}
+for i in range(len(target)):
+    target[i] = t_map[target[i]]
+
+fig, ax = plt.subplots(3, 3, figsize=(14, 14))
+nns = [2, 3, 4, 5, 10, 15, 30, 50, 100]
+i, j = 0, 0
+for n_neighbors in tqdm(nns):
+    fit = umap.UMAP(n_neighbors=n_neighbors)
+    u = fit.fit_transform(geo_arr, y=target)
+    sns.scatterplot(x=u[:,0], y=u[:,1], c=colors, ax=ax[j, i])
+    ax[j, i].set_title(f'n={n_neighbors}')
+    if i < 2: i += 1
+    else: i = 0; j += 1
+
+import umap
+
+fit = umap.UMAP(n_neighbors=50, min_dist=0.5)
+u = fit.fit_transform(geo_arr)
+
+fig = px.scatter(
+    x=u[:,0], y=u[:,1],
+    color=geo['continent'],
+    custom_data=[geo['country'], geo['city']],
+    color_discrete_sequence=palette
+)
+fig.update_traces(
+    hovertemplate="\n".join([
+        "city: %{customdata[1]}",
+        "country: %{customdata[0]}"
+    ])
+)
+
+fig.write_html("umap-earth-2d.html", include_plotlyjs="cdn", full_html=False)
+
+import pandas as pd
+
+umapped = pd.DataFrame({
+    'x': u[:,0],
+    'y': u[:,1],
+    'continent': geo['continent'],
+    'country': geo['country'],
+    'city': geo['city']
+})
+
+umapped.to_csv('umapped.csv', sep='|', index=False)
+
+from sklearn.decomposition import PCA
+
+pca = PCA(n_components=2)  # this means we will create 2-d space
+p = pca.fit_transform(geo_arr)
+fig = px.scatter(
+    x=p[:,0], y=p[:,1],
+    color=geo['continent'],
+    custom_data=[geo['country'], geo['city']],
+    color_discrete_sequence=palette
+)
+fig.update_traces(
+    hovertemplate="\n".join([
+        "city: %{customdata[1]}",
+        "country: %{customdata[0]}"
+    ])
+)
+
+fig.write_html("pca-earth-2d.html", include_plotlyjs="cdn", full_html=False)
+
+
+
+        
+        
+
+
+
+