-
Notifications
You must be signed in to change notification settings - Fork 35
/
streamlit_app.py
249 lines (207 loc) · 11 KB
/
streamlit_app.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
import streamlit as st
import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestRegressor
from sklearn.metrics import mean_squared_error, r2_score
import altair as alt
import time
import zipfile
# Page title
st.set_page_config(page_title='ML model builder', page_icon='🏗️')
st.title('🏗️ ML model builder')
with st.expander('About this app'):
st.markdown('**What can this app do?**')
st.info('This app allow users to build a machine learning (ML) model in an end-to-end workflow. Particularly, this encompasses data upload, data pre-processing, ML model building and post-model analysis.')
st.markdown('**How to use the app?**')
st.warning('To engage with the app, go to the sidebar and 1. Select a data set and 2. Adjust the model parameters by adjusting the various slider widgets. As a result, this would initiate the ML model building process, display the model results as well as allowing users to download the generated models and accompanying data.')
st.markdown('**Under the hood**')
st.markdown('Data sets:')
st.code('''- Drug solubility data set
''', language='markdown')
st.markdown('Libraries used:')
st.code('''- Pandas for data wrangling
- Scikit-learn for building a machine learning model
- Altair for chart creation
- Streamlit for user interface
''', language='markdown')
# Sidebar for accepting input parameters
with st.sidebar:
# Load data
st.header('1.1. Input data')
st.markdown('**1. Use custom data**')
uploaded_file = st.file_uploader("Upload a CSV file", type=["csv"])
if uploaded_file is not None:
df = pd.read_csv(uploaded_file, index_col=False)
# Download example data
@st.cache_data
def convert_df(input_df):
return input_df.to_csv(index=False).encode('utf-8')
example_csv = pd.read_csv('https://raw.githubusercontent.com/dataprofessor/data/master/delaney_solubility_with_descriptors.csv')
csv = convert_df(example_csv)
st.download_button(
label="Download example CSV",
data=csv,
file_name='delaney_solubility_with_descriptors.csv',
mime='text/csv',
)
# Select example data
st.markdown('**1.2. Use example data**')
example_data = st.toggle('Load example data')
if example_data:
df = pd.read_csv('https://raw.githubusercontent.com/dataprofessor/data/master/delaney_solubility_with_descriptors.csv')
st.header('2. Set Parameters')
parameter_split_size = st.slider('Data split ratio (% for Training Set)', 10, 90, 80, 5)
st.subheader('2.1. Learning Parameters')
with st.expander('See parameters'):
parameter_n_estimators = st.slider('Number of estimators (n_estimators)', 0, 1000, 100, 100)
parameter_max_features = st.select_slider('Max features (max_features)', options=['all', 'sqrt', 'log2'])
parameter_min_samples_split = st.slider('Minimum number of samples required to split an internal node (min_samples_split)', 2, 10, 2, 1)
parameter_min_samples_leaf = st.slider('Minimum number of samples required to be at a leaf node (min_samples_leaf)', 1, 10, 2, 1)
st.subheader('2.2. General Parameters')
with st.expander('See parameters', expanded=False):
parameter_random_state = st.slider('Seed number (random_state)', 0, 1000, 42, 1)
parameter_criterion = st.select_slider('Performance measure (criterion)', options=['squared_error', 'absolute_error', 'friedman_mse'])
parameter_bootstrap = st.select_slider('Bootstrap samples when building trees (bootstrap)', options=[True, False])
parameter_oob_score = st.select_slider('Whether to use out-of-bag samples to estimate the R^2 on unseen data (oob_score)', options=[False, True])
sleep_time = st.slider('Sleep time', 0, 3, 0)
# Initiate the model building process
if uploaded_file or example_data:
with st.status("Running ...", expanded=True) as status:
st.write("Loading data ...")
time.sleep(sleep_time)
st.write("Preparing data ...")
time.sleep(sleep_time)
X = df.iloc[:,:-1]
y = df.iloc[:,-1]
st.write("Splitting data ...")
time.sleep(sleep_time)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=(100-parameter_split_size)/100, random_state=parameter_random_state)
st.write("Model training ...")
time.sleep(sleep_time)
if parameter_max_features == 'all':
parameter_max_features = None
parameter_max_features_metric = X.shape[1]
rf = RandomForestRegressor(
n_estimators=parameter_n_estimators,
max_features=parameter_max_features,
min_samples_split=parameter_min_samples_split,
min_samples_leaf=parameter_min_samples_leaf,
random_state=parameter_random_state,
criterion=parameter_criterion,
bootstrap=parameter_bootstrap,
oob_score=parameter_oob_score)
rf.fit(X_train, y_train)
st.write("Applying model to make predictions ...")
time.sleep(sleep_time)
y_train_pred = rf.predict(X_train)
y_test_pred = rf.predict(X_test)
st.write("Evaluating performance metrics ...")
time.sleep(sleep_time)
train_mse = mean_squared_error(y_train, y_train_pred)
train_r2 = r2_score(y_train, y_train_pred)
test_mse = mean_squared_error(y_test, y_test_pred)
test_r2 = r2_score(y_test, y_test_pred)
st.write("Displaying performance metrics ...")
time.sleep(sleep_time)
parameter_criterion_string = ' '.join([x.capitalize() for x in parameter_criterion.split('_')])
#if 'Mse' in parameter_criterion_string:
# parameter_criterion_string = parameter_criterion_string.replace('Mse', 'MSE')
rf_results = pd.DataFrame(['Random forest', train_mse, train_r2, test_mse, test_r2]).transpose()
rf_results.columns = ['Method', f'Training {parameter_criterion_string}', 'Training R2', f'Test {parameter_criterion_string}', 'Test R2']
# Convert objects to numerics
for col in rf_results.columns:
rf_results[col] = pd.to_numeric(rf_results[col], errors='ignore')
# Round to 3 digits
rf_results = rf_results.round(3)
status.update(label="Status", state="complete", expanded=False)
# Display data info
st.header('Input data', divider='rainbow')
col = st.columns(4)
col[0].metric(label="No. of samples", value=X.shape[0], delta="")
col[1].metric(label="No. of X variables", value=X.shape[1], delta="")
col[2].metric(label="No. of Training samples", value=X_train.shape[0], delta="")
col[3].metric(label="No. of Test samples", value=X_test.shape[0], delta="")
with st.expander('Initial dataset', expanded=True):
st.dataframe(df, height=210, use_container_width=True)
with st.expander('Train split', expanded=False):
train_col = st.columns((3,1))
with train_col[0]:
st.markdown('**X**')
st.dataframe(X_train, height=210, hide_index=True, use_container_width=True)
with train_col[1]:
st.markdown('**y**')
st.dataframe(y_train, height=210, hide_index=True, use_container_width=True)
with st.expander('Test split', expanded=False):
test_col = st.columns((3,1))
with test_col[0]:
st.markdown('**X**')
st.dataframe(X_test, height=210, hide_index=True, use_container_width=True)
with test_col[1]:
st.markdown('**y**')
st.dataframe(y_test, height=210, hide_index=True, use_container_width=True)
# Zip dataset files
df.to_csv('dataset.csv', index=False)
X_train.to_csv('X_train.csv', index=False)
y_train.to_csv('y_train.csv', index=False)
X_test.to_csv('X_test.csv', index=False)
y_test.to_csv('y_test.csv', index=False)
list_files = ['dataset.csv', 'X_train.csv', 'y_train.csv', 'X_test.csv', 'y_test.csv']
with zipfile.ZipFile('dataset.zip', 'w') as zipF:
for file in list_files:
zipF.write(file, compress_type=zipfile.ZIP_DEFLATED)
with open('dataset.zip', 'rb') as datazip:
btn = st.download_button(
label='Download ZIP',
data=datazip,
file_name="dataset.zip",
mime="application/octet-stream"
)
# Display model parameters
st.header('Model parameters', divider='rainbow')
parameters_col = st.columns(3)
parameters_col[0].metric(label="Data split ratio (% for Training Set)", value=parameter_split_size, delta="")
parameters_col[1].metric(label="Number of estimators (n_estimators)", value=parameter_n_estimators, delta="")
parameters_col[2].metric(label="Max features (max_features)", value=parameter_max_features_metric, delta="")
# Display feature importance plot
importances = rf.feature_importances_
feature_names = list(X.columns)
forest_importances = pd.Series(importances, index=feature_names)
df_importance = forest_importances.reset_index().rename(columns={'index': 'feature', 0: 'value'})
bars = alt.Chart(df_importance).mark_bar(size=40).encode(
x='value:Q',
y=alt.Y('feature:N', sort='-x')
).properties(height=250)
performance_col = st.columns((2, 0.2, 3))
with performance_col[0]:
st.header('Model performance', divider='rainbow')
st.dataframe(rf_results.T.reset_index().rename(columns={'index': 'Parameter', 0: 'Value'}))
with performance_col[2]:
st.header('Feature importance', divider='rainbow')
st.altair_chart(bars, theme='streamlit', use_container_width=True)
# Prediction results
st.header('Prediction results', divider='rainbow')
s_y_train = pd.Series(y_train, name='actual').reset_index(drop=True)
s_y_train_pred = pd.Series(y_train_pred, name='predicted').reset_index(drop=True)
df_train = pd.DataFrame(data=[s_y_train, s_y_train_pred], index=None).T
df_train['class'] = 'train'
s_y_test = pd.Series(y_test, name='actual').reset_index(drop=True)
s_y_test_pred = pd.Series(y_test_pred, name='predicted').reset_index(drop=True)
df_test = pd.DataFrame(data=[s_y_test, s_y_test_pred], index=None).T
df_test['class'] = 'test'
df_prediction = pd.concat([df_train, df_test], axis=0)
prediction_col = st.columns((2, 0.2, 3))
# Display dataframe
with prediction_col[0]:
st.dataframe(df_prediction, height=320, use_container_width=True)
# Display scatter plot of actual vs predicted values
with prediction_col[2]:
scatter = alt.Chart(df_prediction).mark_circle(size=60).encode(
x='actual',
y='predicted',
color='class'
)
st.altair_chart(scatter, theme='streamlit', use_container_width=True)
# Ask for CSV upload if none is detected
else:
st.warning('👈 Upload a CSV file or click *"Load example data"* to get started!')