### Run Shiny Applications from GitHub

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Provides commands to directly run various Shiny applications from GitHub without local code download. Each command launches a specific interactive application.

```R
# 1. 主应用：Lasso/Ridge/Elastic Net 模型精度比较（含面试题全部解答）
shiny::runGitHub('englianhu/binary.com-interview-question')
```

```R
# 2. Q1App：自动采集汇价 → 建模预测 → 绘制K线+预测图
shiny::runGitHub('englianhu/binary.com-interview-question', subdir = 'Q1')
# 用户选择货币对和预测日期后，系统自动拟合 glmnet 模型并以 highcharter 绘图
```

```R
# 3. testRealTimeTransc：实时交易系统（采集→预测→自动下单→结算→盈亏展示）
shiny::runGitHub('englianhu/binary.com-interview-question', subdir = '实验室/testRealTimeTransc')
# 使用 TFX 包获取 TrueFX 实时汇价，cronR 定时触发 GARCH 预测与下单
```

```R
# 4. Q1App2：庄家（Banker）与赌徒（Punter）双角色金融投注模拟
shiny::runGitHub('englianhu/binary.com-interview-question', subdir = '实验室/Q1App2')
```

```R
# 5. Q2App：双变量泊松模型排队论（运筹学）模拟
shiny::runGitHub('englianhu/binary.com-interview-question', subdir = 'Q2')
```

--------------------------------

### Monte Carlo Simulation: Kelly vs. Optimal f Distribution

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Compares the Kelly criterion with the optimal f distribution using Monte Carlo simulations. Requires functions optKelly and opt.f to be defined elsewhere.

```R
compare <- data.frame(kelly = numeric(5000), opt.f = numeric(5000))
for (i in 1:5000) {
  compare$kelly[i] <- optKelly(2, -0.5, 0.55, 500, 1)$opt.kelly
  compare$opt.f[i] <- opt.f(2,   -0.5, 0.55, 500, 1)$opt.f
}
ggplot(compare) + 
  stat_density(aes(x = kelly, colour = 'Kelly'),  adjust = 2, fill = NA, geom = 'line') + 
  stat_density(aes(x = opt.f, colour = 'Opt.f'), adjust = 2, fill = NA, geom = 'line') + 
  labs(title = '凯利标准 vs 最优 f 分布比较', x = '最优仓位比例', colour = '方法')
```

--------------------------------

### optKelly() / opt.f() — Kelly Criterion and Optimal f Bet Sizing

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Implements the Kelly Criterion and Optimal f (Ralph Vince) capital management algorithms. Calculates the cumulative log-equity across different bet sizes (f ∈ [0, lev]) using Monte Carlo simulation. Returns the optimal position size and a complete results matrix, which can be used with ggplot2 to plot distribution graphs for comparison.

```APIDOC
## optKelly() / opt.f() — 凯利标准与最优 f 投注比例

实现凯利标准（Kelly Criterion）和最优 f（Ralph Vince）两种资金管理算法，通过蒙特卡洛模拟在不同投注比例 f ∈ [0, lev] 上计算累积对数盈亏，返回最优仓位比例及完整结果矩阵，可用 ggplot2 绘制分布图比较两者差异。

```r
source('函数/kelly.opt.R')
library(ggplot2)

# 参数：盈利倍率=2，亏损=-0.5，胜率=0.55，模拟500次，最大杠杆=1
result_kelly <- optKelly(win = 2, loss = -0.5, p = 0.55, obs = 500, lev = 1)
cat('最优凯利比例:', result_kelly$opt.kelly, '\n')
# 最优凯利比例: 0.366

result_optf  <- opt.f(win = 2, loss = -0.5, p = 0.55, obs = 500, lev = 1)
cat('最优 f 比例:', result_optf$opt.f, '\n')
# 最优 f 比例: 0.340
```
```

--------------------------------

### simGarch() — Univariate GARCH Model Rolling Forecast Simulation

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Performs a day-by-day rolling fit of GARCH-class models (supporting sGARCH, fGARCH, gjrGARCH, iGARCH, csGARCH, realGARCH, etc.) on OHLC data for a specified currency pair. Outputs daily one-step-ahead forecast values compared against actual values as an xts time series. ARMA orders are automatically selected by `armaSearch()`. Supports bootstrap confidence intervals and parallel computation.

```APIDOC
## simGarch() — 单变量 GARCH 模型滚动预测模拟

对指定货币对的 OHLC 数据，逐日滚动拟合 GARCH 类模型（支持 sGARCH、fGARCH、gjrGARCH、iGARCH、csGARCH、realGARCH 等），并输出每日一步超前预测值与实际值的对比 xts 时间序列。ARMA 阶数由 `armaSearch()` 自动选取，支持自助法（bootstrap）置信区间与并行计算。

```r
library(rugarch); library(lubridate); library(xts); library(quantmod)
source('函数/simGarch.R')

# 准备 xts 格式 OHLCV 数据（美日汇率）
getSymbols('JPY=X', from = '2014-01-01', to = '2017-08-31', auto.assign = TRUE)
mbase <- `JPY=X` %>%
  na.omit

# 使用 sGARCH(1,1) + 正态分布，预测日均价（Hi+Lo)/2）
result <- simGarch(
  mbase,
  .solver        = 'hybrid',
  .prCat         = 'Mn',           # 预测 (Hi+Lo)/2 均价
  .baseDate      = ymd('2016-01-01'),
  .maPeriod      = 'years',
  .unit          = 1,
  .variance.model = list(model = 'sGARCH', garchOrder = c(1,1),
                         submodel = NULL, external.regressors = NULL,
                         variance.targeting = FALSE),
  .mean.model     = list(armaOrder = c(1,1), include.mean = TRUE),
  .dist.model     = 'norm'
)

# result 为 xts：含 Point.Forecast（预测）和实际 OHLCV 列
tail(result[, c('Point.Forecast', 'USDJPY.Mn')], 5)
#            Point.Forecast USDJPY.Mn
# 2017-08-25         109.21    109.48
# 2017-08-28         109.35    109.52
# 2017-08-29         109.18    109.17
# 2017-08-30         109.44    109.88
# 2017-08-31         109.60    110.02
```
```

--------------------------------

### Calculate Kelly Bet Staking based on GARCH Predictions

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Calculates optimal bet staking using the Kelly formula based on GARCH model predictions for high/low prices. Filters out trades with negative expected value. Requires the kellyBet.R script.

```R
library(dplyr); library(data.table); library(magrittr)
source('函数/kellyBet.R')

# pred.data 为 simGarch 输出后整理的预测数据框，含以下必要列：
# LatestDate.GMT, Lst.High, Lst.Low, ForecastDate.GMT, Fct.High, Fct.Low, EMprob, COMOdds
pred.data <- data.frame(
  LatestDate.GMT   = as.Date('2017-06-27'),
  Lst.High         = 112.399,
  Lst.Low          = 111.863,
  ForecastDate.GMT = 'T+1',
  Fct.High         = 112.547,
  Fct.Low          = 111.766,
  EMprob           = 0.828,
  COMOdds          = 1.95
)

result <- kellyBet(pred.data, initialFundSize = 10000)
# result 含 ProbB, ProbS, Adv (优势值), Staking (最优投注比例)
# Adv = (EMprob * COMOdds) + ((1-EMprob) * -1)
# Staking = Adv / COMOdds，负值截断为 0
```

--------------------------------

### armaSearch() — Automatic ARMA Order Search

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Automatically searches through all combinations of ARMA(p,1,q) orders (p,q ∈ 0..5) to select the optimal order based on the AIC information criterion. Supports 'CSS-ML', 'ML', and 'CSS' fitting methods and includes automatic fallback retries. Returns a complete data frame of AIC comparisons.

```APIDOC
## armaSearch() — ARMA 阶数自动搜索

遍历 ARIMA(p,1,q)（p,q ∈ 0..5）的所有组合，以 AIC 信息准则选出最优阶数，支持 `CSS-ML`、`ML`、`CSS` 三种拟合方法，并在方法失败时自动降级重试，返回完整的 AIC 比较数据框。

```r
library(magrittr)
source('函数/armaSearch.R')

# 使用 quantmod 下载美日汇率日线数据
library(quantmod)
getSymbols('JPY=X', from = '2016-01-01', to = '2017-12-31', auto.assign = TRUE)
cl_data <- Cl(`JPY=X`) %>%
  na.omit

# 自动搜索最优 ARMA(p,q) 阶数
result <- armaSearch(cl_data, .method = 'CSS-ML')
# 输出示例：method = 'CSS-ML', the min AIC = -2345.12, p = 1, q = 1

# 查看所有 AIC 比较表
head(result[order(result$AIC), ])
#    p q       AIC
# 7  1 1 -2345.12
# 8  1 2 -2343.87
# 2  0 1 -2341.05
```
```

--------------------------------

### simETS() — Exponential Smoothing (ETS) Rolling Forecast Simulation

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Performs rolling one-step-ahead forecasts on currency pair OHLC data using the exponential smoothing state-space model from `forecast::ets()`. The model type can be specified (e.g., 'ZZZ' for automatic selection, or specific types like 'ANN', 'MAM'). Supports simulation paths and bootstrap confidence intervals.

```APIDOC
## simETS() — 指数平滑（ETS）滚动预测模拟

基于 `forecast::ets()` 的指数平滑状态空间模型，对货币对 OHLC 数据进行滚动单步预测，模型类型可指定（`ZZZ` 自动选择，或指定如 `ANN`、`MAM` 等），支持模拟路径与自助法置信区间。

```r
library(forecast); library(lubridate); library(xts); library(quantmod)
source('函数/simETS.R')

getSymbols('JPY=X', from = '2014-01-01', to = '2017-08-31', auto.assign = TRUE)
mbase <- `JPY=X` %>%
  na.omit

# 自动选择 ETS 模型，预测日最高价
result_hi <- simETS(
  mbase,
  .model     = 'ZZZ',        # 自动选择误差/趋势/季节性类型
  .damped    = NULL,
  .prCat     = 'Hi',         # 预测日最高价
  .baseDate  = ymd('2016-01-01'),
  .maPeriod  = 'years',
  .unit      = 1,
  .simulate  = FALSE,
  .bootstrap = FALSE
)

# 对比预测值与实际最高价
tail(result_hi[, c('Point.Forecast', 'JPY.X.High')], 3)
#            Point.Forecast JPY.X.High
# 2017-08-29         109.60     109.72
# 2017-08-30         110.01     110.15
# 2017-08-31         110.22     110.38
```
```

--------------------------------

### Simulate GARCH Model Rolling Forecast

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Performs daily rolling fitting of GARCH-class models (sGARCH, fGARCH, gjrGARCH, etc.) on OHLC data for a specified currency pair. Outputs daily one-step-ahead forecast values against actuals as an xts time series. ARMA order is automatically selected by `armaSearch()`, supporting bootstrap confidence intervals and parallel computation.

```r
library(rugarch); library(lubridate); library(xts); library(quantmod)
source('函数/simGarch.R')

# 准备 xts 格式 OHLCV 数据（美日汇率）
getSymbols('JPY=X', from = '2014-01-01', to = '2017-08-31', auto.assign = TRUE)
mbase <- `JPY=X` %>% na.omit

# 使用 sGARCH(1,1) + 正态分布，预测日均价（Hi+Lo)/2）
result <- simGarch(
  mbase,
  .solver        = 'hybrid',
  .prCat         = 'Mn',           # 预测 (Hi+Lo)/2 均价
  .baseDate      = ymd('2016-01-01'),
  .maPeriod      = 'years',
  .unit          = 1,
  .variance.model = list(model = 'sGARCH', garchOrder = c(1,1),
                         submodel = NULL, external.regressors = NULL,
                         variance.targeting = FALSE),
  .mean.model     = list(armaOrder = c(1,1), include.mean = TRUE),
  .dist.model     = 'norm'
)

# result 为 xts：含 Point.Forecast（预测）和实际 OHLCV 列
tail(result[, c('Point.Forecast', 'USDJPY.Mn')], 5)
#            Point.Forecast USDJPY.Mn
# 2017-08-25         109.21    109.48
# 2017-08-28         109.35    109.52
# 2017-08-29         109.18    109.17
# 2017-08-30         109.44    109.88
# 2017-08-31         109.60    110.02
```

--------------------------------

### Kelly Criterion and Optimal f Bet Sizing

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Implements the Kelly Criterion and Optimal f (Ralph Vince) capital management algorithms. It calculates cumulative log returns across different bet sizes (f ∈ [0, lev]) using Monte Carlo simulation, returning the optimal position size and a complete results matrix. Useful for plotting distribution differences with ggplot2.

```r
source('函数/kelly.opt.R')
library(ggplot2)

# 参数：盈利倍率=2，亏损=-0.5，胜率=0.55，模拟500次，最大杠杆=1
result_kelly <- optKelly(win = 2, loss = -0.5, p = 0.55, obs = 500, lev = 1)
cat('最优凯利比例:', result_kelly$opt.kelly, '\n')
# 最优凯利比例: 0.366

result_optf  <- opt.f(win = 2, loss = -0.5, p = 0.55, obs = 500, lev = 1)
cat('最优 f 比例:', result_optf$opt.f, '\n')
# 最优 f 比例: 0.340
```

--------------------------------

### filterFX() — Forex Data Filtering and Standardization

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Filters and standardizes column names of raw xts forex objects downloaded by `quantmod::getSymbols()`, based on currency code and price type (Open/High/Low/Close/Adjusted). Removes missing values and returns a clean, single-column xts series. Supports seven major currency pairs: AUDUSD, EURUSD, GBPUSD, USDCHF, USDCAD, USDCNY, USDJPY.

```APIDOC
## filterFX() — 外汇数据过滤与标准化

将 `quantmod::getSymbols()` 下载的原始 xts 外汇对象，按货币代码和价格类型（开/高/低/收/调整）进行过滤与列名标准化，去除缺失值并返回干净的单列 xts 序列。支持七种主要货币对：AUDUSD、EURUSD、GBPUSD、USDCHF、USDCAD、USDCNY、USDJPY。

```r
library(quantmod); library(magrittr); library(stringr)
source('函数/filterFX.R')

# 下载美日汇率
getSymbols('JPY=X', from = '2015-01-01', to = '2017-12-31', auto.assign = TRUE)
mbase <- `JPY=X`

# 提取日收盘价并标准化列名
usdjpy_cl <- filterFX(mbase, currency = 'JPY=X', price = 'Cl')
# 列名自动规范为 'USDJPY'，NA 已去除
head(usdjpy_cl)
#            USDJPY
# 2015-01-02 119.63
# 2015-01-05 119.17

# 提取日最高价
usdjpy_hi <- filterFX(mbase, currency = 'JPY=X', price = 'Hi')
head(usdjpy_hi)
#            USDJPY
# 2015-01-02 120.11
# 2015-01-05 119.80
```
```

--------------------------------

### Auto-search ARMA Order using AIC

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Automatically searches through ARIMA(p,1,q) combinations (p,q ∈ 0..5) to find the optimal order based on the AIC information criterion. Supports CSS-ML, ML, and CSS fitting methods, with automatic fallback retries on failure. Returns a complete AIC comparison data frame.

```r
library(magrittr)
source('函数/armaSearch.R')

# 使用 quantmod 下载美日汇率日线数据
library(quantmod)
getSymbols('JPY=X', from = '2016-01-01', to = '2017-12-31', auto.assign = TRUE)
cl_data <- Cl(`JPY=X`) %>% na.omit

# 自动搜索最优 ARMA(p,q) 阶数
result <- armaSearch(cl_data, .method = 'CSS-ML')
# 输出示例：method = 'CSS-ML', the min AIC = -2345.12, p = 1, q = 1

# 查看所有 AIC 比较表
head(result[order(result$AIC), ])
#    p q       AIC
# 7  1 1 -2345.12
# 8  1 2 -2343.87
# 2  0 1 -2341.05
```

--------------------------------

### Elastic Net Forex Price Prediction using glmnet

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Predicts forex closing prices using Elastic Net regression (alpha=0.8) with the glmnet package. The h() function creates feature matrices, and glmPrice() handles model fitting and prediction. Requires sourced R scripts for helper functions.

```R
library(glmnet); library(magrittr); library(dplyr)
source('实验室/shinyFunction/h.R')
source('实验室/shinyFunction/glmPrice.R')
source('实验室/shinyFunction/filterLAD.R')
source('实验室/shinyFunction/loadLAD.R')

# 载入数据
datam  <- loadLAD()
fundDT <- filterLAD(startDate = Sys.Date() - 365, endDate = Sys.Date())$fundDT

# 构造特征矩阵（使用 h2 矩阵方案，预测 daily.mean2）
xy <- h(fundDT, family = 'gaussian', xy.matrix = 'h2',
        setform = 'l4', yv = 'daily.mean2')

# 使用预训练模型（alpha=0.8, lambda.1se）进行预测
fitgaum16.alpha08 <- readRDS('./shinyData/fitgaum16.alpha08.rds')
pd <- predict(fitgaum16.alpha08,
              newx       = xy$x,
              s          = fitgaum16.alpha08$lambda.1se,
              pred.type  = 'class') %>%
  as.data.frame() %>%
  rename(Pred = 1)

# 合并预测值与原始数据
result <- fundDT %>%
  mutate(HL.Mean = (LAD.High + LAD.Low) / 2) %>%
  bind_cols(pd)
tail(result[, c('Date', 'HL.Mean', 'Pred')], 3)
```

--------------------------------

### Filter and Standardize FX Data

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Filters and standardizes raw xts Forex objects downloaded by `quantmod::getSymbols()`. It selects data by currency code and price type (Open/High/Low/Close/Adjusted), removes NA values, and returns a clean single-column xts series. Supports seven major currency pairs: AUDUSD, EURUSD, GBPUSD, USDCHF, USDCAD, USDCNY, USDJPY.

```r
library(quantmod); library(magrittr); library(stringr)
source('函数/filterFX.R')

# 下载美日汇率
getSymbols('JPY=X', from = '2015-01-01', to = '2017-12-31', auto.assign = TRUE)
mbase <- `JPY=X`

# 提取日收盘价并标准化列名
usdjpy_cl <- filterFX(mbase, currency = 'JPY=X', price = 'Cl')
# 列名自动规范为 'USDJPY'，NA 已去除
head(usdjpy_cl)
#            USDJPY
# 2015-01-02 119.63
# 2015-01-05 119.17

# 提取日最高价
usdjpy_hi <- filterFX(mbase, currency = 'JPY=X', price = 'Hi')
head(usdjpy_hi)
#            USDJPY
# 2015-01-02 120.11
# 2015-01-05 119.80
```

--------------------------------

### Simulate ETS Model Rolling Forecast

Source: https://context7.com/englianhu/binary.com-interview-question/llms.txt

Performs rolling one-step-ahead predictions using exponential smoothing state space models from `forecast::ets()`. Model types can be specified or automatically selected (`ZZZ`). Supports simulation paths and bootstrap confidence intervals.

```r
library(forecast); library(lubridate); library(xts); library(quantmod)
source('函数/simETS.R')

getSymbols('JPY=X', from = '2014-01-01', to = '2017-08-31', auto.assign = TRUE)
mbase <- `JPY=X` %>% na.omit

# 自动选择 ETS 模型，预测日最高价
result_hi <- simETS(
  mbase,
  .model     = 'ZZZ',        # 自动选择误差/趋势/季节性类型
  .damped    = NULL,
  .prCat     = 'Hi',         # 预测日最高价
  .baseDate  = ymd('2016-01-01'),
  .maPeriod  = 'years',
  .unit      = 1,
  .simulate  = FALSE,
  .bootstrap = FALSE
)

# 对比预测值与实际最高价
tail(result_hi[, c('Point.Forecast', 'JPY.X.High')], 3)
#            Point.Forecast JPY.X.High
# 2017-08-29         109.60     109.72
# 2017-08-30         110.01     110.15
# 2017-08-31         110.22     110.38
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.