big update

RESEARCH/SelfLearn-Research.R

@@ -5,6 +5,9 @@
 # https://www.udemy.com/self-learning-trading-robot/?couponCode=LAZYTRADE7-10
 # Script to gather financial data, transform it and to perform
 # Supervised Deep Learning Classification Modelling
+# 
+# Purpose of this script is to simulate results in the several possible variants
+# NOTE: delete all previously create models from the folder R_selflearning/model after updating h2o engine
 #
 # load libraries to use and custom functions
 library(tidyverse)
@@ -21,6 +24,7 @@ source("C:/Users/fxtrams/Documents/000_TradingRepo/R_selflearning/self_learn_ai_
 source("C:/Users/fxtrams/Documents/000_TradingRepo/R_selflearning/test_model.R")
 #absolute path to store model objects (useful when scheduling tasks)
 path_model <- "C:/Users/fxtrams/Documents/000_TradingRepo/R_selflearning/model"
+
 h2o.init()
 
 ### Create For loop to test possible outcomes and test those strategies
@@ -75,6 +79,7 @@ files_to_analyse <-list.files(file.path(getwd(),"RESEARCH/"), pattern="*.rds", f
 # e.g.: str_extract(files_to_analyse[2], "(?<=Result-)(.*)(?=.rds)")
 # 
 for (FILE in files_to_analyse) {
+  #FILE <- files_to_analyse[3]
   #extract nbars from the file name, note we use only those of the latest date
   num_bars <- FILE %>% str_extract(paste0("(?<=",Sys.Date(),"-Result-)(.*)(?=.rds)"))
   if(!exists("summary_file")){

_FUN/create_labelled_data.R

@@ -0,0 +1,79 @@
+## FUNCTION create_labelled_data
+## PURPOSE: function gets price data in each column
+## it is splitting this data by periods and transposes the data. 
+## additionally it is label the data based on the simple logic assigning it to 2 categories based on the difference
+## between beginning and end of the vector
+## finally it is stacking all data and joining everything into the table 
+
+## TEST:
+# library(tidyverse)
+# library(lubridate)
+# pathT2 <- "C:/Program Files (x86)/FxPro - Terminal2/MQL4/Files/"
+# prices <- read_csv(file.path(pathT2, "AI_CP1.csv"), col_names = F)
+# prices$X1 <- ymd_hms(prices$X1)
+# write_rds(prices, "test_data/prices.rds")
+
+#' Create labelled data
+#' https://www.udemy.com/self-learning-trading-robot/?couponCode=LAZYTRADE7-10
+#'
+#' @param x - data set containing a table where 1st column is a Time index and other columns containing financial asset price values
+#' @param n - number of rows we intend to split and transpose the data 
+#' @param type - type of the label required. Can be either "classification" or "regression". "classification" will return either "BU" or "BE",
+#' "regression" will return the difference between first value and the last value in each row
+#'
+#' @return function returns transposed data. One column called 'LABEL' indicate achieved value of the label.
+#' Transposed values from every column are stacked one to each other
+#' 
+#' @export
+#'
+#' @examples
+create_labelled_data <- function(x, n = 50, type = "regression"){
+  require(tidyverse)
+  #source("C:/Users/fxtrams/Documents/000_TradingRepo/R_selflearning/_FUN/load_data.R")
+  #x <- load_data(path_terminal = "C:/Program Files (x86)/FxPro - Terminal2/MQL4/Files/", trade_log_file = "AI_CP", time_period = 1, data_deepth = "50000")
+  #x <- read_rds("_TEST_DATA/price_dataset.rds")
+  #n <- 100
+  #type <- "classification"
+  #type <- "regression"
+  #
+  nr <- nrow(x)
+  dat11 <- x %>% 
+    # remove column 1 with data and time information
+    select(-1) %>% 
+    # split dataset into several objects each containing n rows (it will be a list)
+    split(rep(1:ceiling(nr/n), each=n, length.out=nr)) #list
+  # remove last element of the list
+  dat11[length(dat11)] <- NULL
+
+  # operations within the list
+  for (i in 1:length(dat11)) {
+    #i <- 2
+    if(type == "classification"){
+
+        # classify by 2 classes 'BU', 'BE'
+        if(!exists("dfr12")){
+          dfr12 <- dat11[i] %>% as.data.frame() %>% t() %>% as_tibble(.name_repair = "universal") %>% 
+            mutate(LABEL = ifelse(.[[1]]>.[[n]], "BU", "BE"))} else {
+            dfr12 <- dat11[i] %>% as.data.frame() %>% t() %>% as_tibble(.name_repair = "universal") %>% 
+              mutate(LABEL = ifelse(.[[1]]>.[[n]], "BU", "BE")) %>% 
+              bind_rows(dfr12)
+          }
+    } else if(type == "regression"){
+      # add label with numeric difference {in pips}
+      # i <- 1
+      if(!exists("dfr12")){
+        dfr12 <- dat11[i] %>% as.data.frame() %>% t() %>% as_tibble(.name_repair = "universal", verbose =F) %>%
+          mutate(LABEL = 10000*(.[[1]]-.[[n]]))} else {
+          dfr12 <- dat11[i] %>% as.data.frame() %>% t() %>% as_tibble(.name_repair = "universal", verbose =F) %>%
+            mutate(LABEL = 10000*(.[[1]]-.[[n]])) %>% 
+            #oldest data will be on top of the dataframe!
+            bind_rows(dfr12)
+        }
+
+
+    }
+  }
+
+  return(dfr12)
+
+}

_FUN/create_transposed_data.R

@@ -0,0 +1,51 @@
+## FUNCTION create_transposed_data
+## PURPOSE: function gets indicator data in each column
+## it is splitting this data by periods and transpose the data. 
+## additionally it is label the data based on the simple logic assigning it to 2 categories based on the difference
+## between beginning and end of the vector
+## finally it is stacking all data and joining everything into the table 
+
+## TEST:
+# library(tidyverse)
+# library(lubridate)
+# pathT2 <- "C:/Program Files (x86)/FxPro - Terminal2/MQL4/Files/"
+# macd <- read_csv(file.path(pathT2, "AI_Macd1.csv"), col_names = F)
+# macd$X1 <- ymd_hms(macd$X1)
+# write_rds(macd, "test_data/macd.rds")
+
+#' Create Transposed Data
+#' https://www.udemy.com/self-learning-trading-robot/?couponCode=LAZYTRADE7-10
+#'
+#' @param x - data set containing a table where 1st column is a Time index and other columns containing financial asset indicator values
+#' @param n - number of rows we intend to split and transpose the data 
+#'
+#' @return function returns transposed data. Transposed values from every column are stacked one to each other
+#' @export
+#'
+#' @examples
+#' 
+create_transposed_data <- function(x, n = 50){
+  require(tidyverse)
+  #source("C:/Users/fxtrams/Documents/000_TradingRepo/R_selflearning/_FUN/load_data.R")
+  #n <- 100
+  #x <- load_data(path_terminal = "C:/Program Files (x86)/FxPro - Terminal2/MQL4/Files/", trade_log_file = "AI_Macd", time_period = 1, data_deepth = "50000")
+  #x <- read_rds("_TEST_DATA/indicator_dataset.rds")
+  nr <- nrow(x)
+  dat11 <- x %>% select(-1) %>% split(rep(1:ceiling(nr/n), each=n, length.out=nr)) #list
+  dat11[length(dat11)] <- NULL
+
+  # operations within the list
+  for (i in 1:length(dat11)) {
+    #i <- 1
+
+    if(!exists("dfr12")){
+      dfr12 <- dat11[i] %>% as.data.frame() %>% t() %>% as_tibble(.name_repair = "universal", verbose =F) } else {
+        dfr12 <- dat11[i] %>% as.data.frame() %>% t() %>% as_tibble(.name_repair = "universal", verbose =F) %>% bind_rows(dfr12)
+      }
+
+  }
+
+  return(dfr12)
+
+}
+

_FUN/load_data.R

@@ -0,0 +1,79 @@
+# import data function
+# (C) 2018 Vladimir Zhbanko
+# -------------------------
+# Import Data to R
+# -------------------------
+#' Load Data Function
+#' https://www.udemy.com/self-learning-trading-robot/?couponCode=LAZYTRADE7-10
+#' 
+#' Function imports file and change date column type. Function return the dataframe with trade data. 
+#' -> update: data_deepth argument was added to optimize code and separate data needed to train and score the model
+#'
+#' @param path_terminal - path to the MT4 terminal, string
+#' @param trade_log_file - csv file name where the data is stored, without ".csv"
+#' @param time_period - data periodicity in minutes, can be 1, 15, 60
+#' @param data_deepth - collected data deepth in rows. describe how many rows in original file to read
+#'
+#' @return - dataframe with asset data in columns where X1 column is in a POSIXct format
+#' @export
+#'
+#' @examples  
+#' prices <- load_data(path_terminal = "C:/Program Files (x86)/FxPro - Terminal2/MQL4/Files/",
+#'                     trade_log_file = "AI_CP", 
+#'                     time_period = 1,
+#'                     data_deepth = "14200")
+#' 
+load_data <- function(path_terminal, trade_log_file, time_period = 1, data_deepth = 14200){
+  # path_terminal <- "C:/Program Files (x86)/FxPro - Terminal2/MQL4/Files/"
+  # path_terminal <- file.path(getwd(), "test_data")
+  # trade_log_file <- "AI_CP"
+  # trade_log_file <- "AI_Macd"
+  # time_period <- 1
+  # data_deepth <- "100000"
+  require(tidyverse)
+  require(lubridate)
+  DFT1 <- try(read_csv(file = file.path(path_terminal, paste0(trade_log_file, time_period, "-", data_deepth, ".csv")),
+                       col_names = F),
+              silent = TRUE)
+  if(class(DFT1)[1] == "try-error") {stop("Error reading file. File with trades may not exist yet!",
+                                       call. = FALSE)}
+  if(!nrow(DFT1)==0){
+    # data frame preparation
+    DFT1$X1 <- ymd_hms(DFT1$X1)
+    if(trade_log_file == "AI_CP"){
+    ## divide JPY pairs by 100
+    DFT2 <- DFT1[ , c(8,10,18,22,24,25,26)]/100
+    DFT3 <- DFT1[, -c(8,10,18,22,24,25,26)] %>% bind_cols(DFT2) %>% select(X1,X2,X3,X4,X5,X6,X7,X8,
+                                                                           X9,X10,X11,X12,X13,X14,X15,
+                                                                           X16,X17,X18,X19,X20,X21,X22,
+                                                                           X23,X24,X25,X26,X27,X28,X29)
+    return(DFT3)
+    }
+
+    return(DFT1)
+  } else {
+    stop("Data log is empty!", call. = FALSE)
+    }
+
+}
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+