ium_464953/sacred/sacred_model_creator.py
Michal Gulczynski af91b85a30 ium_07 sacred
2024-06-11 19:06:13 +02:00

109 lines
4.7 KiB
Python

import pandas as pd
import os
import numpy as np
from kaggle.api.kaggle_api_extended import KaggleApi
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import LabelEncoder
import pickle
from sacred import Experiment
from sacred.observers import MongoObserver, FileObserver
# Tworzenie eksperymentu
ex = Experiment('123456') # Zastąp '123456' swoim numerem indeksu
# Dodanie obserwatorów
ex.observers.append(MongoObserver(url='mongodb://admin:IUM_2021@tzietkiewicz.vm.wmi.amu.edu.pl:27017'))
ex.observers.append(FileObserver('my_experiment_logs'))
def download_dataset(dataset_address, destination_folder):
api = KaggleApi()
api.authenticate()
api.dataset_download_files(dataset_address, path=destination_folder, unzip=True)
def check_datasets_presence():
dataset_1 = "Spotify_Dataset.csv"
dataset_2 = "spotify_songs.csv"
destination_folder = "datasets"
if not os.path.exists(destination_folder):
os.makedirs(destination_folder)
print(f"Utworzono folder: {destination_folder}")
else:
print(f"Folder {destination_folder} już istnieje.")
if dataset_1 not in os.listdir(destination_folder):
download_dataset('gulczas/spotify-dataset', destination_folder)
if dataset_2 not in os.listdir(destination_folder):
download_dataset('joebeachcapital/30000-spotify-songs', destination_folder)
def datasets_preparation():
df_1 = pd.read_csv("datasets/spotify_songs.csv")
df_2 = pd.read_csv("datasets/Spotify_Dataset.csv", sep=";")
df_1 = df_1.dropna()
df_2 = df_2.dropna()
df_2 = df_2.rename(columns={'Title': 'track_name'})
columns_to_remove_df_1 = ['track_id', 'track_album_id', 'track_album_name', 'track_album_release_date',
'playlist_id', 'playlist_subgenre']
columns_to_remove_df_2 = ['Date','# of Artist', 'Artist (Ind.)', '# of Nationality',
'Nationality', 'Continent', 'Points (Total)',
'Points (Ind for each Artist/Nat)', 'id', 'Song URL']
df_1 = df_1.drop(columns=columns_to_remove_df_1)
df_2 = df_2.drop(columns=columns_to_remove_df_2)
df_1 = df_1.drop_duplicates(subset=['track_name'])
df_2 = df_2.drop_duplicates(subset=['track_name'])
le = LabelEncoder()
unique_names_df2 = df_2['track_name'].unique()
diff_df = df_1[~df_1['track_name'].isin(unique_names_df2)]
diff_df = diff_df.iloc[:10000]
diff_df['track_artist'] = le.fit_transform(diff_df.track_artist)
diff_df['playlist_name'] = le.fit_transform(diff_df.playlist_name)
diff_df['playlist_genre'] = le.fit_transform(diff_df.playlist_genre)
if "docker_test_dataset.csv" not in os.listdir("datasets"):
diff_df.to_csv("datasets/docker_test_dataset.csv", index=False)
result_df = pd.merge(df_1, df_2, on='track_name', how='inner')
result_df = result_df.drop_duplicates(subset=['track_name'])
columns_to_remove_result_df = ['Rank', 'Artists', 'Danceability', 'Energy', 'Loudness',
'Speechiness', 'Acousticness', 'Instrumentalness', 'Valence']
result_df = result_df.drop(columns=columns_to_remove_result_df)
result_df['track_artist'] = le.fit_transform(result_df.track_artist)
result_df['playlist_name'] = le.fit_transform(result_df.playlist_name)
result_df['playlist_genre'] = le.fit_transform(result_df.playlist_genre)
return result_df
@ex.config
def config():
test_size = 0.10
random_state = 42
model_filename = 'model.pkl'
@ex.main
def run_experiment(test_size, random_state, model_filename):
check_datasets_presence()
result_df = datasets_preparation()
Y = result_df[['playlist_genre']]
X = result_df.drop(columns='playlist_genre')
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=test_size, random_state=random_state)
Y_train = np.ravel(Y_train)
Y_test = np.ravel(Y_test)
scaler = StandardScaler()
numeric_columns = X_train.select_dtypes(include=['int', 'float']).columns
X_train_scaled = scaler.fit_transform(X_train[numeric_columns])
X_test_scaled = scaler.transform(X_test[numeric_columns])
model = LogisticRegression(max_iter=1000)
model.fit(X_train_scaled, Y_train)
Y_pred = model.predict(X_test_scaled)
accuracy = accuracy_score(Y_test, Y_pred)
ex.log_scalar('accuracy', accuracy)
if os.path.exists(model_filename):
os.remove(model_filename)
with open(model_filename, 'wb') as file:
pickle.dump(model, file)
ex.add_artifact(model_filename)
ex.add_resource(__file__)
print("Accuracy:", accuracy)
return accuracy
if __name__ == '__main__':
ex.run_commandline()