-
Upload 10 files
-
based on melisekm/natural-disasters-from-social-media
based on "melisekm/natural-disasters-from-social-media"
labels made by featurizing via `cu_cat` and taking first 2 pc via `k-means`
```
from cu_cat._dep_manager import deps
deps.cuml=None
AA = TableVectorizer().fit_transform(dataset['train']['label'])
pca = PCA(2)
pca_data = pd.DataFrame(pca.fit_transform(AA),columns=['PC1','PC2'])
kmeans =KMeans(n_clusters=2).fit(AA)
pca_data['cluster'] = pd.Categorical(kmeans.labels_)
fig,ax = plt.subplots()
scatter = ax.scatter(pca_data['PC1'], pca_data['PC2'],c=pca_data['cluster'],cmap='Set3',alpha=0.7)
legend1 = ax.legend(*scatter.legend_elements(),
loc="upper left", title="")
ax.add_artist(legend1)
plt.savefig('test')
dataset = dataset.map(lambda examples: {'label': TableVectorizer().fit_transform(examples['label'])}, batched=True)
dataset = dataset.map(lambda examples: {'label': KMeans(n_clusters=2).fit(examples['label']).labels_}, batched=True)
```
