Datasets:
Sébastien De Greef
commited on
Commit
•
1e1b538
1
Parent(s):
491a75a
Add scripts to create indicators and sequences, and download crypto data
Browse files- create_indicators.py +40 -0
- create_sequences.py +64 -0
- crypto_data.py +69 -0
- download.py +12 -9
create_indicators.py
ADDED
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import pandas as pd
|
2 |
+
import numpy as np
|
3 |
+
|
4 |
+
df = pd.read_csv('candles.csv')
|
5 |
+
|
6 |
+
|
7 |
+
from talib import RSI, BBANDS, MACD, ATR, EMA, SMA
|
8 |
+
|
9 |
+
|
10 |
+
# group by market
|
11 |
+
grouped = df.groupby('market')
|
12 |
+
|
13 |
+
# for each market calculate the indicators and add them to the dataframe
|
14 |
+
for market, group in grouped:
|
15 |
+
# calculate the indicators
|
16 |
+
print('Calculating indicators for', market)
|
17 |
+
df.loc[group.index,'rsi'] = RSI(group['close'], timeperiod=14)
|
18 |
+
|
19 |
+
upper, middle, lower = BBANDS(group['close'], timeperiod=20)
|
20 |
+
df.loc[group.index,'bb_upper'] = upper
|
21 |
+
df.loc[group.index,'bb_middle'] = middle
|
22 |
+
df.loc[group.index,'bb_lower'] = lower
|
23 |
+
|
24 |
+
macd, macdsignal, macdhist = MACD(group['close'], fastperiod=12, slowperiod=26, signalperiod=9)
|
25 |
+
df.loc[group.index,'macd'] = macd
|
26 |
+
df.loc[group.index,'macdsignal'] = macdsignal
|
27 |
+
df.loc[group.index,'macdhist'] = macdhist
|
28 |
+
|
29 |
+
df.loc[group.index,'atr'] = ATR(group['high'], group['low'], group['close'], timeperiod=14)
|
30 |
+
|
31 |
+
df.loc[group.index,'ema'] = EMA(group['close'], timeperiod=30)
|
32 |
+
df.loc[group.index,'sma'] = SMA(group['close'], timeperiod=30)
|
33 |
+
|
34 |
+
# drop the rows with NaN values
|
35 |
+
df = df.dropna()
|
36 |
+
|
37 |
+
|
38 |
+
# save the dataframe to a new file
|
39 |
+
print(df.tail())
|
40 |
+
df.to_csv('indicators.csv', index=False)
|
create_sequences.py
ADDED
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
|
3 |
+
# data = data[['market', 'time', 'open', 'high', 'low', 'close', 'volume']]
|
4 |
+
# transform this dataset so to 'market', 'start', 'column', 'value1', 'value2', 'value3', 'value(n)'
|
5 |
+
import pandas as pd
|
6 |
+
import numpy as np
|
7 |
+
from tqdm import tqdm
|
8 |
+
|
9 |
+
columns = ['open','close','volume', 'rsi', 'sma']
|
10 |
+
window_size = 10
|
11 |
+
|
12 |
+
def create_sequences(df, columns=columns, window_size=window_size):
|
13 |
+
# group by market
|
14 |
+
grouped = df.groupby('market')
|
15 |
+
|
16 |
+
# create a list of dataframes
|
17 |
+
dfs = []
|
18 |
+
|
19 |
+
for name, group in tqdm(grouped):
|
20 |
+
# create a new dataframe
|
21 |
+
new_df = pd.DataFrame()
|
22 |
+
new_df['market'] = name
|
23 |
+
# create a list of lists
|
24 |
+
sequences = []
|
25 |
+
# only include the close column
|
26 |
+
# iterate over the rows of the dataframe
|
27 |
+
for i in range(len(group) - window_size):
|
28 |
+
# create a sequence
|
29 |
+
sequence = group.iloc[i:i+window_size][columns].values
|
30 |
+
# transpose the sequence so that it is a column
|
31 |
+
sequence = sequence.T
|
32 |
+
|
33 |
+
# create a dataframe from the sequence
|
34 |
+
sequence = pd.DataFrame(sequence)
|
35 |
+
|
36 |
+
# add the market, time, column_name to the sequence
|
37 |
+
sequence['market'] = name
|
38 |
+
sequence['time'] = group.iloc[i+window_size]['time']
|
39 |
+
sequence['column'] = columns
|
40 |
+
|
41 |
+
# set market, time as the first columns and index
|
42 |
+
sequence = sequence.set_index(['market', 'time', 'column'])
|
43 |
+
|
44 |
+
# add the sequence to the list of sequences
|
45 |
+
sequences.append(sequence)
|
46 |
+
if len(sequences) == 0:
|
47 |
+
continue
|
48 |
+
# create a dataframe from the list of lists
|
49 |
+
new_df = pd.concat(sequences)
|
50 |
+
# add the dataframe to the list of dataframes
|
51 |
+
dfs.append(new_df)
|
52 |
+
|
53 |
+
# concatenate the list of dataframes
|
54 |
+
final_df = pd.concat(dfs)
|
55 |
+
|
56 |
+
return final_df
|
57 |
+
|
58 |
+
df = pd.read_csv('indicators.csv')
|
59 |
+
|
60 |
+
# create the sequences
|
61 |
+
sequences = create_sequences(df, columns=columns, window_size=15)
|
62 |
+
|
63 |
+
# save the sequences to a new file
|
64 |
+
sequences.to_csv('sequences.csv')
|
crypto_data.py
ADDED
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from datasets import DatasetBuilder, DownloadManager, DatasetInfo, BuilderConfig, SplitGenerator, Split, Features, Value
|
2 |
+
import pandas as pd
|
3 |
+
|
4 |
+
# Define custom configurations for the dataset
|
5 |
+
class CryptoDataConfig(BuilderConfig):
|
6 |
+
def __init__(self, features, **kwargs):
|
7 |
+
super().__init__(**kwargs)
|
8 |
+
self.features = features
|
9 |
+
|
10 |
+
class CryptoDataDataset(DatasetBuilder):
|
11 |
+
# Define different dataset configurations here
|
12 |
+
BUILDER_CONFIGS = [
|
13 |
+
CryptoDataConfig(
|
14 |
+
name="candles",
|
15 |
+
description="This configuration includes open, high, low, close, and volume.",
|
16 |
+
features=Features({
|
17 |
+
"date": Value("string"),
|
18 |
+
"open": Value("float"),
|
19 |
+
"high": Value("float"),
|
20 |
+
"low": Value("float"),
|
21 |
+
"close": Value("float"),
|
22 |
+
"volume": Value("float")
|
23 |
+
})
|
24 |
+
),
|
25 |
+
CryptoDataConfig(
|
26 |
+
name="indicators",
|
27 |
+
description="This configuration extends basic CryptoDatas with RSI, SMA, and EMA indicators.",
|
28 |
+
features=Features({
|
29 |
+
"date": Value("string"),
|
30 |
+
"open": Value("float"),
|
31 |
+
"high": Value("float"),
|
32 |
+
"low": Value("float"),
|
33 |
+
"close": Value("float"),
|
34 |
+
"volume": Value("float"),
|
35 |
+
"rsi": Value("float"),
|
36 |
+
"sma": Value("float"),
|
37 |
+
"ema": Value("float")
|
38 |
+
})
|
39 |
+
),
|
40 |
+
]
|
41 |
+
|
42 |
+
def _info(self):
|
43 |
+
return DatasetInfo(
|
44 |
+
description=f"CryptoData dataset for {self.config.name}",
|
45 |
+
features=self.config.features,
|
46 |
+
supervised_keys=None,
|
47 |
+
homepage="https://hub.huggingface.co/datasets/sebdg/crypto_data",
|
48 |
+
citation="No citation for this dataset."
|
49 |
+
)
|
50 |
+
|
51 |
+
def _split_generators(self, dl_manager: DownloadManager):
|
52 |
+
# Here, you can define how to split your dataset (e.g., into training, validation, test)
|
53 |
+
# This example assumes a single CSV file without predefined splits.
|
54 |
+
# You can modify this method if you have different needs.
|
55 |
+
return [
|
56 |
+
SplitGenerator(
|
57 |
+
name=Split.TRAIN,
|
58 |
+
gen_kwargs={"filepath": "indicators.csv"},
|
59 |
+
),
|
60 |
+
]
|
61 |
+
|
62 |
+
def _generate_examples(self, filepath):
|
63 |
+
# Here, we open the provided CSV file and yield each row as a single example.
|
64 |
+
with open(filepath, encoding="utf-8") as csv_file:
|
65 |
+
data = pd.read_csv(csv_file)
|
66 |
+
for id, row in data.iterrows():
|
67 |
+
# Select features based on the dataset configuration
|
68 |
+
features = {feature: row[feature] for feature in self.config.features if feature in row}
|
69 |
+
yield id, features
|
download.py
CHANGED
@@ -41,15 +41,18 @@ print('Data downloaded and saved to assets.csv and markets.csv')
|
|
41 |
if not os.path.exists('candles'):
|
42 |
os.makedirs('candles')
|
43 |
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
#
|
50 |
-
|
51 |
-
|
52 |
-
#
|
|
|
|
|
|
|
53 |
|
54 |
# print('Ticker data downloaded')
|
55 |
|
|
|
41 |
if not os.path.exists('candles'):
|
42 |
os.makedirs('candles')
|
43 |
|
44 |
+
for market in markets['market']:
|
45 |
+
print('Downloading', market)
|
46 |
+
url = f'https://api.bitvavo.com/v2/{market}/candles?interval=1d&limit=1440'
|
47 |
+
response = requests.get(url)
|
48 |
+
data = response.json()
|
49 |
+
#print(data)
|
50 |
+
data = pd.DataFrame(data, columns=['time', 'open', 'high', 'low', 'close', 'volume'])
|
51 |
+
data['market'] = market
|
52 |
+
# set market as the first column
|
53 |
+
data = data[['market', 'time', 'open', 'high', 'low', 'close', 'volume']]
|
54 |
+
data.to_csv(f'candles/{market}.csv', index=False)
|
55 |
+
time.sleep(0.5)
|
56 |
|
57 |
# print('Ticker data downloaded')
|
58 |
|