Replace AffinityPropagation with SpectralClustering
This commit is contained in:
parent
7e356fed37
commit
b4e90c1174
|
@ -5,7 +5,7 @@ from matplotlib.pyplot import *
|
|||
from pandas import DataFrame
|
||||
from seaborn import heatmap, set_style, set_theme, pairplot
|
||||
from sklearn.metrics import silhouette_score, calinski_harabasz_score
|
||||
from sklearn.cluster import KMeans, Birch, AffinityPropagation, MeanShift, DBSCAN
|
||||
from sklearn.cluster import KMeans, Birch, SpectralClustering, MeanShift, DBSCAN
|
||||
|
||||
from preprocessing import parse_data, filter_dataframe
|
||||
|
||||
|
@ -15,8 +15,8 @@ def choose_model(model):
|
|||
return KMeans(random_state=42)
|
||||
elif model == "birch":
|
||||
return Birch()
|
||||
elif model == "affinity":
|
||||
return AffinityPropagation(random_state=42)
|
||||
elif model == "spectral":
|
||||
return SpectralClustering()
|
||||
elif model == "meanshift":
|
||||
return MeanShift()
|
||||
elif model == "dbscan":
|
||||
|
@ -29,6 +29,7 @@ def predict_data(data, model, results, sample):
|
|||
start_time = time.time()
|
||||
prediction = model.fit_predict(data)
|
||||
execution_time = time.time() - start_time
|
||||
cluster_number = len(set(prediction))
|
||||
calinski = calinski_harabasz_score(X=data, labels=prediction)
|
||||
silhouette = silhouette_score(
|
||||
X=data,
|
||||
|
@ -41,7 +42,7 @@ def predict_data(data, model, results, sample):
|
|||
df=results,
|
||||
model=model_name,
|
||||
prediction=prediction,
|
||||
clusters=len(set(prediction)),
|
||||
clusters=cluster_number,
|
||||
calinski=calinski,
|
||||
silhouette=silhouette,
|
||||
time=execution_time,
|
||||
|
@ -79,17 +80,12 @@ def plot_scatter_plot(results):
|
|||
fig.savefig(f"docs/assets/{fig_title.replace(' ', '_').lower()}.png")
|
||||
|
||||
|
||||
def print_dataframe(df):
|
||||
df.set_index("model")
|
||||
print(df)
|
||||
|
||||
|
||||
def show_results(results):
|
||||
set_theme()
|
||||
set_style("white")
|
||||
plot_heatmap(results=results)
|
||||
plot_scatter_plot(results=results)
|
||||
print_dataframe(df=results)
|
||||
print(results)
|
||||
|
||||
|
||||
def create_result_dataframes():
|
||||
|
@ -103,8 +99,7 @@ def create_result_dataframes():
|
|||
"time",
|
||||
]
|
||||
)
|
||||
indexed_results = results.set_index("model")
|
||||
return indexed_results, indexed_results
|
||||
return results, results
|
||||
|
||||
|
||||
def populate_results(df, model, clusters, prediction, calinski, silhouette, time):
|
||||
|
@ -153,7 +148,7 @@ def usage():
|
|||
|
||||
|
||||
def main():
|
||||
models = ["kmeans", "birch", "affinity", "meanshift", "dbscan"]
|
||||
models = ["kmeans", "birch", "spectral", "meanshift", "dbscan"]
|
||||
if len(argv) != 4:
|
||||
usage()
|
||||
case, sample = argv[2], int(argv[3])
|
||||
|
@ -172,7 +167,7 @@ def main():
|
|||
individual_result.append(model_results)
|
||||
)
|
||||
complete_results.set_index("model")
|
||||
print_dataframe(df=complete_results)
|
||||
print(complete_results)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
|
|
Loading…
Reference in New Issue