New release. Lots of stuff.

README.md

@@ -10,7 +10,7 @@ This repository is organized as it follows:
 <li> Implement the Toda-Yamamoto procedure to test for Granger-causality between correlated cryptocoins.</li>
 <li> Train and test SOTA machine learning models to forecast cryptocoin price series (namely GRU, LSTM, CatBoost, LightGBM and XGBoost).</li>
 </ul>
-<li><b>Data:</b> pre-built datasets adopted in the above-mentioned analyses, spanning 27 months from 13/03/2020 to 22/06/2022. </li>
+<li><b>Data:</b> pre-built datasets adopted in the above-mentioned analyses, spanning 33 months from 20-02-2020 to 26-02-2023. </li>
 
 ## Data
 The data sources used to gather information about cryptocurrency trends are [CoinMarketCap](https://www.coinmarketcap.com/) and [Binance](https://www.binance.com/).

analysis/causality/toda_yamamoto.R

@@ -1,6 +1,6 @@
 # File: toda_yamamoto.R
 # Description: Implementation of the Toda-Yamamoto (1995) procedure to test for Granger causality.
-# File Created: 01/07/2022
+# File Created: 01/02/2023
 # R Version: 3.6
 
 
@@ -14,26 +14,30 @@ library(zoo)
 library(tseries)
 
 # File properties
-data_path <- './data/datasets/binance.csv'
+wd <- getwd()
+data_path <- file.path(wd, 'data/binance_notest.csv')
 
 # Set seed
-set.seed(123)
+set.seed(1234)
 
 # Load data
-data <- read_delim(data_path, delim = '\t')
-coins <- names(data[, 2:16])
+data <- read_delim(data_path, delim = ',')
+coins <- names(data[, 2:17])
 
 # Transform into daily observations
 daily_freq <- as.Date(cut(data$Date, 'day'))
-new_data <- aggregate(ADA ~ daily_freq, data, mean)
-colnames(new_data) <- c('Date', 'ADA')
+new_data <- aggregate(DOGE ~ daily_freq, data, mean)
+colnames(new_data) <- c('Date', 'XLM')
 
-for (i in coins[2:15])
+for (i in coins[2:16])
 { coin <- aggregate(data[[i]] ~ daily_freq, data, mean)
   colnames(coin) <- c('Date', i)
   new_data <- cbind(new_data, coin[2])
 }
 
+# Write output
+sink('results/output.txt')
+
 # ADF test
 for (i in coins)
   { print(i)
@@ -65,7 +69,7 @@ for (i in coins)
 # 1st order differentiation eliminates the unit root, then the maximum order of integration is 1.
 
 # Set up VAR select using all the possible coin pairs to find optimal lag (lower AIC)
-coins_subset <- coins[-c(3, 6)]
+coins_subset <- coins[-c(3, 5)]
 btc_lags <- c()
 eth_lags <- c()
 
@@ -95,11 +99,6 @@ V <- VAR(new_data[c('BTC', i)], p=lag+1, type='both')
 waldtest1 <- wald.test(b=coef(V$varresult[[1]]), Sigma=vcov(V$varresult[[1]]), Terms=c(seq(2, by=2, length=lag)))
 cat(i, 'does not Granger-cause BTC', '\n')
 print(waldtest1$result)
-
-# Wald-test 2 (H0: BTC does not Granger-cause alt-coin)
-waldtest2 <- wald.test(b=coef(V$varresult[[2]]), Sigma=vcov(V$varresult[[2]]), Terms=c(seq(1, by=2, length=lag)))
-cat('BTC does not Granger-cause', i, '\n')
-print(waldtest2$result)
 }
 
 # ETH
@@ -113,10 +112,8 @@ V <- VAR(new_data[c('ETH', i)], p=lag+1, type='both')
 # Wald-test 1 (H0: alt-coin does not Granger-cause ETH)
 waldtest1 <- wald.test(b=coef(V$varresult[[1]]), Sigma=vcov(V$varresult[[1]]), Terms=c(seq(2, by=2, length=lag)))
 cat(i, 'does not Granger-cause ETH', '\n')
-print(waldtest1$result) 
-
-# Wald-test 2 (H0: ETH does not Granger-cause alt-coin)
-waldtest2 <- wald.test(b=coef(V$varresult[[2]]), Sigma=vcov(V$varresult[[2]]), Terms=c(seq(1, by=2, length=lag)))
-cat('ETH does not Granger-cause', i, '\n')
-print(waldtest2$result)
+print(waldtest1$result)
 }
+
+# Stop writing output
+sink()

analysis/correlation/ohlc/correlations_daily.py

@@ -1,25 +1,27 @@
 """
 File: correlations_daily.py
-Description: Correlation Analysis with daily granularity (source: coinmarketcap dataset).
-File Created: 06/01/2022
+Description: Correlation Analysis with daily granularity (source: raw.csv dataset).
+File Created: 01/01/2023
 Python Version: 3.9
 """
 
 # Imports
+import sys
 import os
 import numpy as np
 import pandas as pd
 
 # File properties
-data_path = './data/datasets/coinmarketcap.csv'
-corr_data_path = './analysis/correlation/correlogram_data'
+root_dir = sys.path[1]
+data_path = os.path.join(root_dir, 'data/datasets/coinmarketcap.csv')
+corr_data_path = os.path.join(root_dir, 'correlation/correlogram_data')
 column_names = ['coin', 'x', 'y', 'radius', 'arc_begin', 'arc_end', 'color']
 arc_begin = 0
 arc_end = 360
 color = 1
 
 # Extract btc and eth benchmarks
-data = pd.read_csv(data_path, sep='\t')
+data = pd.read_csv(data_path, sep=',')
 data['Date'] = pd.to_datetime(data['Date'])
 data.drop(['Market Cap', 'Volume'], axis=1, inplace=True)
 data['Avg OHLC Price'] = data[['Open', 'High', 'Low', 'Close']].mean(axis=1)
@@ -86,5 +88,5 @@
 corr_data_eth = pd.concat(list_data_eth)
 file_name_btc = os.path.join(corr_data_path, 'daily_btc_OHLC.csv')
 file_name_eth = os.path.join(corr_data_path, 'daily_eth_OHLC.csv')
-corr_data_btc.to_csv(file_name_btc, sep='\t', encoding='utf-8', index=False)
-corr_data_eth.to_csv(file_name_eth, sep='\t', encoding='utf-8', index=False)
+corr_data_btc.to_csv(file_name_btc, sep=',', encoding='utf-8', index=False)
+corr_data_eth.to_csv(file_name_eth, sep=',', encoding='utf-8', index=False)

analysis/correlation/ohlc/correlations_monthly_sw.py

@@ -1,26 +1,28 @@
 """
 File: correlations_monthly_sw.py
-Description: Correlation Analysis with monthly granularity and sliding window (source: coinmarketcap dataset).
-File Created: 06/01/2022
+Description: Correlation Analysis with monthly granularity and sliding window (source: raw.csv dataset).
+File Created: 01/01/2023
 Python Version: 3.9
 """
 
 # Imports
+import sys
 import os
 import numpy as np
 import pandas as pd
 
 # File properties
-data_path = './data/datasets/coinmarketcap.csv'
-corr_data_path = './analysis/correlation/correlogram_data'
+root_dir = sys.path[1]
+data_path = os.path.join(root_dir, 'data/datasets/coinmarketcap.csv')
+corr_data_path = os.path.join(root_dir, 'correlation/correlogram_data')
 window_size = '30D'
 column_names = ['coin', 'x', 'y', 'radius', 'arc_begin', 'arc_end', 'color']
 arc_begin = 0
 arc_end = 360
 color = 1
 
 # Extract btc and eth benchmarks
-data = pd.read_csv(data_path, sep='\t')
+data = pd.read_csv(data_path, sep=',')
 data['Date'] = pd.to_datetime(data['Date'])
 data.drop(['Market Cap', 'Volume'], axis=1, inplace=True)
 data['Avg OHLC Price'] = data[['Open', 'High', 'Low', 'Close']].mean(axis=1)
@@ -87,5 +89,5 @@
 corr_data_eth = pd.concat(list_data_eth)
 file_name_btc = os.path.join(corr_data_path, 'monthly_sw_btc_OHLC.csv')
 file_name_eth = os.path.join(corr_data_path, 'monthly_sw_eth_OHLC.csv')
-corr_data_btc.to_csv(file_name_btc, sep='\t', encoding='utf-8', index=False)
-corr_data_eth.to_csv(file_name_eth, sep='\t', encoding='utf-8', index=False)
+corr_data_btc.to_csv(file_name_btc, sep=',', encoding='utf-8', index=False)
+corr_data_eth.to_csv(file_name_eth, sep=',', encoding='utf-8', index=False)

analysis/correlation/ohlc/correlations_monthly_tw.py

@@ -1,26 +1,28 @@
 """
 File: correlations_monthly_tw.py
-Description: Correlation Analysis with monthly granularity and thumbing window (source: coinmarketcap dataset).
-File Created: 06/01/2022
+Description: Correlation Analysis with monthly granularity and thumbing window (source: raw.csv dataset).
+File Created: 01/01/2023
 Python Version: 3.9
 """
 
 # Imports
+import sys
 import os
 import numpy as np
 import pandas as pd
 
 # File properties
-data_path = './data/datasets/coinmarketcap.csv'
-corr_data_path = './analysis/correlation/correlogram_data'
+root_dir = sys.path[1]
+data_path = os.path.join(root_dir, 'data/datasets/coinmarketcap.csv')
+corr_data_path = os.path.join(root_dir, 'correlation/correlogram_data')
 time_frame = 'M'
 column_names = ['coin', 'x', 'y', 'radius', 'arc_begin', 'arc_end', 'color']
 arc_begin = 0
 arc_end = 360
 color = 1
 
 # Extract btc and eth benchmarks
-data = pd.read_csv(data_path, sep='\t')
+data = pd.read_csv(data_path, sep=',')
 data['Date'] = pd.to_datetime(data['Date'])
 data.drop(['Market Cap', 'Volume'], axis=1, inplace=True)
 data['Avg OHLC Price'] = data[['Open', 'High', 'Low', 'Close']].mean(axis=1)
@@ -87,5 +89,5 @@
 corr_data_eth = pd.concat(list_data_eth)
 file_name_btc = os.path.join(corr_data_path, 'monthly_tw_btc_OHLC.csv')
 file_name_eth = os.path.join(corr_data_path, 'monthly_tw_eth_OHLC.csv')
-corr_data_btc.to_csv(file_name_btc, sep='\t', encoding='utf-8', index=False)
-corr_data_eth.to_csv(file_name_eth, sep='\t', encoding='utf-8', index=False)
+corr_data_btc.to_csv(file_name_btc, sep=',', encoding='utf-8', index=False)
+corr_data_eth.to_csv(file_name_eth, sep=',', encoding='utf-8', index=False)

analysis/correlation/ohlc/correlations_weekly_sw.py

@@ -1,26 +1,28 @@
 """
 File: correlations_weekly_sw.py
-Description: Correlation Analysis with weekly granularity and sliding window (source: coinmarketcap dataset).
-File Created: 06/01/2022
+Description: Correlation Analysis with weekly granularity and sliding window (source: raw.csv dataset).
+File Created: 01/01/2023
 Python Version: 3.9
 """
 
 # Imports
+import sys
 import os
 import numpy as np
 import pandas as pd
 
 # File properties
-data_path = './data/datasets/coinmarketcap.csv'
-corr_data_path = './analysis/correlation/correlogram_data'
+root_dir = sys.path[1]
+data_path = os.path.join(root_dir, 'data/datasets/coinmarketcap.csv')
+corr_data_path = os.path.join(root_dir, 'correlation/correlogram_data')
 window_size = '7D'
 column_names = ['coin', 'x', 'y', 'radius', 'arc_begin', 'arc_end', 'color']
 arc_begin = 0
 arc_end = 360
 color = 1
 
 # Extract btc and eth benchmarks
-data = pd.read_csv(data_path, sep='\t')
+data = pd.read_csv(data_path, sep=',')
 data['Date'] = pd.to_datetime(data['Date'])
 data.drop(['Market Cap', 'Volume'], axis=1, inplace=True)
 data['Avg OHLC Price'] = data[['Open', 'High', 'Low', 'Close']].mean(axis=1)
@@ -87,5 +89,5 @@
 corr_data_eth = pd.concat(list_data_eth)
 file_name_btc = os.path.join(corr_data_path, 'weekly_sw_btc_OHLC.csv')
 file_name_eth = os.path.join(corr_data_path, 'weekly_sw_eth_OHLC.csv')
-corr_data_btc.to_csv(file_name_btc, sep='\t', encoding='utf-8', index=False)
-corr_data_eth.to_csv(file_name_eth, sep='\t', encoding='utf-8', index=False)
+corr_data_btc.to_csv(file_name_btc, sep=',', encoding='utf-8', index=False)
+corr_data_eth.to_csv(file_name_eth, sep=',', encoding='utf-8', index=False)

analysis/correlation/ohlc/correlations_weekly_tw.py

@@ -1,26 +1,28 @@
 """
 File: correlations_weekly_tw.py
-Description: Correlation Analysis with weekly granularity and thumbing window (source: coinmarketcap dataset).
-File Created: 06/01/2022
+Description: Correlation Analysis with weekly granularity and thumbing window (source: raw.csv dataset).
+File Created: 01/01/2023
 Python Version: 3.9
 """
 
 # Imports
+import sys
 import os
 import numpy as np
 import pandas as pd
 
 # File properties
-data_path = './data/datasets/coinmarketcap.csv'
-corr_data_path = './analysis/correlation/correlogram_data'
+root_dir = sys.path[1]
+data_path = os.path.join(root_dir, 'data/datasets/coinmarketcap.csv')
+corr_data_path = os.path.join(root_dir, 'correlation/correlogram_data')
 time_frame = 'W'
 column_names = ['coin', 'x', 'y', 'radius', 'arc_begin', 'arc_end', 'color']
 arc_begin = 0
 arc_end = 360
 color = 1
 
 # Extract btc and eth benchmarks
-data = pd.read_csv(data_path, sep='\t')
+data = pd.read_csv(data_path, sep=',')
 data['Date'] = pd.to_datetime(data['Date'])
 data.drop(['Market Cap', 'Volume'], axis=1, inplace=True)
 data['Avg OHLC Price'] = data[['Open', 'High', 'Low', 'Close']].mean(axis=1)
@@ -87,5 +89,5 @@
 corr_data_eth = pd.concat(list_data_eth)
 file_name_btc = os.path.join(corr_data_path, 'weekly_tw_btc_OHLC.csv')
 file_name_eth = os.path.join(corr_data_path, 'weekly_tw_eth_OHLC.csv')
-corr_data_btc.to_csv(file_name_btc, sep='\t', encoding='utf-8', index=False)
-corr_data_eth.to_csv(file_name_eth, sep='\t', encoding='utf-8', index=False)
+corr_data_btc.to_csv(file_name_btc, sep=',', encoding='utf-8', index=False)
+corr_data_eth.to_csv(file_name_eth, sep=',', encoding='utf-8', index=False)

analysis/correlation/volume/correlations_daily.py

@@ -1,25 +1,27 @@
 """
 File: correlations_daily.py
-Description: Correlation Analysis with daily granularity (source: coinmarketcap dataset).
-File Created: 06/01/2022
+Description: Correlation Analysis with daily granularity (source: raw.csv dataset).
+File Created: 01/01/2023
 Python Version: 3.9
 """
 
 # Imports
+import sys
 import os
 import numpy as np
 import pandas as pd
 
 # File properties
-data_path = './data/datasets/coinmarketcap.csv'
-corr_data_path = './analysis/correlation/correlogram_data'
+root_dir = sys.path[1]
+data_path = os.path.join(root_dir, 'data/datasets/coinmarketcap.csv')
+corr_data_path = os.path.join(root_dir, 'correlation/correlogram_data')
 column_names = ['coin', 'x', 'y', 'radius', 'arc_begin', 'arc_end', 'color']
 arc_begin = 0
 arc_end = 360
 color = 1
 
 # Extract btc and eth benchmarks
-data = pd.read_csv(data_path, sep='\t')
+data = pd.read_csv(data_path, sep=',')
 data['Date'] = pd.to_datetime(data['Date'])
 data.drop(['Market Cap', 'Open', 'High', 'Low', 'Close'], axis=1, inplace=True)
 group = data.groupby('Coin')
@@ -84,5 +86,5 @@
 corr_data_eth = pd.concat(list_data_eth)
 file_name_btc = os.path.join(corr_data_path, 'daily_btc_VOL.csv')
 file_name_eth = os.path.join(corr_data_path, 'daily_eth_VOL.csv')
-corr_data_btc.to_csv(file_name_btc, sep='\t', encoding='utf-8', index=False)
-corr_data_eth.to_csv(file_name_eth, sep='\t', encoding='utf-8', index=False)
+corr_data_btc.to_csv(file_name_btc, sep=',', encoding='utf-8', index=False)
+corr_data_eth.to_csv(file_name_eth, sep=',', encoding='utf-8', index=False)

analysis/correlation/volume/correlations_monthly_sw.py

@@ -1,26 +1,28 @@
 """
 File: correlations_monthly_sw.py
-Description: Correlation Analysis with monthly granularity and sliding window (source: coinmarketcap dataset).
-File Created: 06/01/2022
+Description: Correlation Analysis with monthly granularity and sliding window (source: raw.csv dataset).
+File Created: 01/01/2023
 Python Version: 3.9
 """
 
 # Imports
+import sys
 import os
 import numpy as np
 import pandas as pd
 
 # File properties
-data_path = './data/datasets/coinmarketcap.csv'
-corr_data_path = './analysis/correlation/correlogram_data'
+root_dir = sys.path[1]
+data_path = os.path.join(root_dir, 'data/datasets/coinmarketcap.csv')
+corr_data_path = os.path.join(root_dir, 'correlation/correlogram_data')
 window_size = '30D'
 column_names = ['coin', 'x', 'y', 'radius', 'arc_begin', 'arc_end', 'color']
 arc_begin = 0
 arc_end = 360
 color = 1
 
 # Extract btc and eth benchmarks
-data = pd.read_csv(data_path, sep='\t')
+data = pd.read_csv(data_path, sep=',')
 data['Date'] = pd.to_datetime(data['Date'])
 data.drop(['Market Cap', 'Open', 'High', 'Low', 'Close'], axis=1, inplace=True)
 group = data.groupby('Coin')
@@ -85,5 +87,5 @@
 corr_data_eth = pd.concat(list_data_eth)
 file_name_btc = os.path.join(corr_data_path, 'monthly_sw_btc_VOL.csv')
 file_name_eth = os.path.join(corr_data_path, 'monthly_sw_eth_VOL.csv')
-corr_data_btc.to_csv(file_name_btc, sep='\t', encoding='utf-8', index=False)
-corr_data_eth.to_csv(file_name_eth, sep='\t', encoding='utf-8', index=False)
+corr_data_btc.to_csv(file_name_btc, sep=',', encoding='utf-8', index=False)
+corr_data_eth.to_csv(file_name_eth, sep=',', encoding='utf-8', index=False)