The goal of YahooTickers is to easily get stock index constituents from
Yahoo Finance without having to worry
about complicated for loops. This may be useful if you wish to focus
as quickly as possible in time-series modeling, but also convenient if
data is needed for it’s side effects (e.g. plotting).
The package depends heavly on quantmod, rsample, sweep and forecast projects.
You can install YahooTickers from github with:
# install.packages("devtools")
devtools::install_github("Reckziegel/YahooTickers")To download all the stock constituents of the Dow Jones Industrial
Average use get_tickers() and get_stocks() in one pipeline.
dow <- get_tickers(dow) %>%
get_stocks(., periodicity = "monthly")
dow
#> # A tibble: 4,209 x 8
#> date tickers open high low close volume adjusted
#> <date> <fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 2010-01-01 MMM 83.1 85.2 79.1 80.5 75208100 58.1
#> 2 2010-02-01 MMM 80.8 81.8 77.2 80.2 75020400 57.8
#> 3 2010-03-01 MMM 80.6 84.5 80.3 83.6 91066100 60.7
#> 4 2010-04-01 MMM 83.9 90.2 82.7 88.7 96407000 64.4
#> 5 2010-05-01 MMM 89.2 90.5 69.0 79.3 109573600 57.6
#> 6 2010-06-01 MMM 78.7 83 72.7 79.0 114407500 57.7
#> 7 2010-07-01 MMM 79.1 87.5 77.0 85.5 89556700 62.5
#> 8 2010-08-01 MMM 86.8 88.4 78.4 78.6 74721100 57.4
#> 9 2010-09-01 MMM 79.5 88 79.3 86.7 64059700 63.8
#> 10 2010-10-01 MMM 87.4 91.5 83.8 84.2 82038100 61.9
#> # ... with 4,199 more rowsWith additional functions provided in the package it’s possible to fit a
model in a rolling or expanding time window. This is done by
get_models().
library(forecast)
merval_models <- get_tickers(merval) %>%
get_stocks(
tickers = .,
periodicity = "monthly",
quiet = TRUE
) %>%
get_returns(
.tbl = .,
.group = tickers,
.type = log,
.omit_na = TRUE,
ret_adj = adjusted # change col names on the fly like dplyr!
) %>%
get_models(
.tbl = .,
.group = tickers,
.col = ret_adj,
.initial = 100,
.assess = 1,
.cumulative = TRUE,
.fun = ets
)
merval_models
#> # A tibble: 672 x 17
#> date tickers data term estimate model.desc sigma logLik AIC BIC
#> <date> <fct> <list> <fct> <dbl> <fct> <dbl> <dbl> <dbl> <dbl>
#> 1 2018-06-01 GGAL.BA <tibbl~ alpha 0.000100 ETS(A,N,N) 0.126 -21.9 49.9 57.7
#> 2 2018-06-01 GGAL.BA <tibbl~ l 0.0397 ETS(A,N,N) 0.126 -21.9 49.9 57.7
#> 3 2018-07-01 GGAL.BA <tibbl~ alpha 0.000100 ETS(A,N,N) 0.126 -22.8 51.6 59.4
#> 4 2018-07-01 GGAL.BA <tibbl~ l 0.0383 ETS(A,N,N) 0.126 -22.8 51.6 59.4
#> 5 2018-08-01 GGAL.BA <tibbl~ alpha 0.000100 ETS(A,N,N) 0.125 -23.0 52.0 59.9
#> 6 2018-08-01 GGAL.BA <tibbl~ l 0.0382 ETS(A,N,N) 0.125 -23.0 52.0 59.9
#> 7 2018-09-01 GGAL.BA <tibbl~ alpha 0.000100 ETS(A,N,N) 0.126 -24.1 54.2 62.1
#> 8 2018-09-01 GGAL.BA <tibbl~ l 0.0366 ETS(A,N,N) 0.126 -24.1 54.2 62.1
#> 9 2018-10-01 GGAL.BA <tibbl~ alpha 0.000100 ETS(A,N,N) 0.126 -25.2 56.4 64.4
#> 10 2018-10-01 GGAL.BA <tibbl~ l 0.0382 ETS(A,N,N) 0.126 -25.2 56.4 64.4
#> # ... with 662 more rows, and 7 more variables: ME <dbl>, RMSE <dbl>,
#> # MAE <dbl>, MPE <dbl>, MAPE <dbl>, MASE <dbl>, ACF1 <dbl>In this case, the the
ETS function
was used to fit the first 100 observations on a expanding window,
.cumulative = TRUE. The output contains the fitted parameters,
selection criterias, as some traditional in-sample error metrics.
To extract the true out-of-sample forecasts use get_forecasts(). This
function automatically identifies the .assess argument in
get_models() and uses it as a target for the forecast horizon.
library(forecast)
merval_models %>%
get_forecasts(.)
#> # A tibble: 336 x 8
#> date tickers ret_adj point_forecast lo.80 lo.95 hi.80 hi.95
#> <date> <fct> <dbl> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 2018-06-01 GGAL.BA -0.0998 0.0397 -0.122 -0.207 0.201 0.286
#> 2 2018-07-01 GGAL.BA 0.0244 0.0383 -0.123 -0.209 0.200 0.285
#> 3 2018-08-01 GGAL.BA -0.123 0.0382 -0.122 -0.207 0.199 0.284
#> 4 2018-09-01 GGAL.BA 0.205 0.0366 -0.125 -0.210 0.198 0.283
#> 5 2018-10-01 GGAL.BA -0.238 0.0382 -0.124 -0.209 0.200 0.286
#> 6 2018-11-01 GGAL.BA 0.202 0.0356 -0.129 -0.216 0.200 0.287
#> 7 2018-12-01 GGAL.BA 0.0136 0.0372 -0.128 -0.215 0.202 0.290
#> 8 2019-01-01 GGAL.BA 0.270 0.0369 -0.127 -0.215 0.201 0.288
#> 9 2019-02-01 GGAL.BA -0.114 0.0391 -0.127 -0.215 0.205 0.293
#> 10 2019-03-01 GGAL.BA -0.0873 0.0377 -0.129 -0.217 0.204 0.292
#> # ... with 326 more rowsCurrently, YahooTickers supports the following stock indexes:
dow: Dow Jones Industrial Average (United States)sp500: S&P 500 (United States)nasdaq: NASDAQ Composite Index (United States)nyse: NYSE Composite Index (United States)amex: AMEX Composite Index (United States)russell2000: Russell 2000 Index (United States)ftse100: FTSE 100 Index (United Kingdom)dax: Deutsche Boerse AG German Stock (Germany)cac40: CAC 40 Index (France)bel20: BEL 20 Index (Belgium)topix: TOPIX Core 30 Index (Japan)hangseng: HangSeng Composite Index (Hong Kong)sensex: S&P BSE SENSEX Index (India)jakarta: Jakarta Stock Exchange Stock Index (Indonesia)bursa: Bursa Malaysia Bhd (Malaysia)nzx50: S&P/NZX 50 Gross Index (New Zealand)kospi: Korea Stock Exchange KOSPI Index (south Korea)taiex: Taiwan Stock Exchange Weighted Index (Taiwan)tsx: S&P/TSX Composite Index (Canada)ibovespa: Ibovespa Brasil Sao Paulo Stock Exchange Index (Brazil)ipc: S&P/BMV IPC (Mexico)ipsa: S&P/CLX IPSA (Chile)merval: Buenos Aires Stock Exchange Merval Index (Argentina)
The Yahoo Finance API works reasonably
well but there are cases in which it does not cover historical data for
all stocks in an index universe. The bel20 it’s a good example:
despite officially having 20 components, Yahoo Finance only maintains
data for 4 stocks. Thus, it is recommended to always check the data
after downloading it!