{"id":380223,"date":"2023-11-22T01:04:19","date_gmt":"2023-11-22T07:04:19","guid":{"rendered":"http:\/\/datageeek.com\/?p=7119"},"modified":"2023-11-22T01:04:19","modified_gmt":"2023-11-22T07:04:19","slug":"time-series-analysis-and-understanding-a-eurobond-fund-with-prophet","status":"publish","type":"post","link":"https:\/\/www.r-bloggers.com\/2023\/11\/time-series-analysis-and-understanding-a-eurobond-fund-with-prophet\/","title":{"rendered":"Time Series Analysis and Understanding a Eurobond Fund with Prophet"},"content":{"rendered":"\r\n\r\n
\r\n[This article was first published on DataGeeek<\/a><\/strong>, and kindly contributed to R-bloggers<\/a>]. (You can report issue about the content on this page here<\/a>)\r\n
Want to share your content on R-bloggers? click here<\/a> if you have a blog, or here<\/a> if you don't.\r\n<\/div>\n\n

The Eurobonds are so popular in Turkey nowadays. The reason for this is that the CDS of Turkey fell for quite a while since the new economic outlook began, according to the economists. We will look into the factors that shape this trend in this article.<\/p>\n\n\n\n

First, we build our data set with the components that we think are related to the Turkish Eurobonds.<\/p>\n\n\n

\nlibrary(tidyverse)\nlibrary(tidymodels)\nlibrary(timetk)\nlibrary(modeltime)\nlibrary(modelStudio)\n\n#GPA(Eurobond Debt Instruments (FX) Fund of a private Turkish bank)\ndf_gpa <- \n  read_csv("https:\/\/raw.githubusercontent.com\/mesdi\/blog\/main\/gpa.csv")\n\ndf_gpa_tidy <- \n  df_gpa %>% \n  janitor::clean_names() %>% \n  mutate(date = parse_date(date, "%m\/%d\/%Y")) %>% \n  select(date, gpa = price)\n\n#CDS(Turkey CDS 5 Years USD)\ndf_cds <- \n  read_csv("https:\/\/raw.githubusercontent.com\/mesdi\/blog\/main\/cds.csv")\n\ndf_cds_tidy <- \n  df_cds %>% \n  janitor::clean_names() %>% \n  mutate(date = parse_date(date, "%m\/%d\/%Y")) %>% \n  select(date, cds = price)\n\n#USDTRY(US Dollar Turkish Lira)\ndf_usdtry <- \n  read_csv("https:\/\/raw.githubusercontent.com\/mesdi\/blog\/main\/usdtry.csv")\n\ndf_usdtry_tidy <- \n  df_usdtry %>% \n  janitor::clean_names() %>% \n  mutate(date = parse_date(date, "%m\/%d\/%Y")) %>% \n  select(date, usdtry = price)\n\n#Federal Funds Effective Rate\ndf_fedfunds <- \n  read_csv("https:\/\/raw.githubusercontent.com\/mesdi\/blog\/main\/fedfunds.csv")\n\n##Converting Fed Funds data from monthly to weekly \nlibrary(padr)\ndf_fedfunds_tidy <- \n  df_fedfunds %>% \n  janitor::clean_names() %>% \n  thicken("week") %>% \n  pad_by_time(date_week, .by = "week") %>%\n  mutate(across(fedfunds, .fns = \\(x) ts_impute_vec(x, period = 1))) %>% \n  select(date = date_week, fedfunds)\n\n#Transforming the variables to PPP (Purchasing power parities) \n#to see the changes(%) relative the last point\ndf_tidy <- \n  df_all %>%\n  mutate(gpa = (gpa\/first(gpa)*100) %>% round(2),\n         cds = (cds\/first(cds)*100) %>% round(2),\n         usdtry = (usdtry\/first(usdtry)*100) %>% round(2),\n         fedfunds = (fedfunds\/first(fedfunds)*100) %>% round(2))\n<\/pre>\n\n\n
Now that we have created our data set, we can examine the anomalies<\/a><\/em> of the variables.<\/p>\n\n\n
\n#Anomaly\ndf_all %>% \n  pivot_longer(cols = -date, names_to = "vars") %>%\n  plot_anomaly_diagnostics(date, \n                           value,\n                           .interactive = FALSE,\n                           .facet_vars = vars,\n                           .facet_ncol = 2)\n<\/pre>\n\n\n
\"\"<\/a><\/figure>\n\n\n\n
We can see that all variables but cds<\/code><\/em> have had a lot of anomalies in the last few years. So we will choose the prophet<\/a><\/em> model which is robust to outliers. Before moving on to our modeling, we will analyze the relations between all variables with the clustering<\/a><\/em> method.<\/p>\n\n\n
\n#Clustering\ndf_tsfeature <- \n  df_all %>% \n  pivot_longer(cols = -date, names_to = "id") %>%\n  group_by(id) %>%\n  tk_tsfeatures(\n    .date_var = date,\n    .value    = value,\n    .period   = 52,\n    .features = c("frequency", "stl_features", "entropy", "acf_features", "mean"),\n    .scale    = TRUE,\n    .prefix   = "ts_"\n  ) %>%\n  ungroup()\n\nset.seed(123)\ntibble(\n  cluster = df_tsfeature %>% \n    select(-id) %>%\n    as.matrix() %>%\n    kmeans(centers = 3, nstart = 100) %>%\n    pluck("cluster")\n) %>%\n  bind_cols(\n    df_tsfeature\n  ) %>% \n  select(cluster, id) %>%\n  right_join(df_all %>% pivot_longer(cols = -date, names_to = "id") , \n             by = "id") %>%\n  group_by(id) %>%\n  plot_time_series(\n    date, value, \n    .color_var   = cluster, \n    .facet_ncol  = 2, \n    .interactive = FALSE,\n    .title = ""\n  ) +\n  theme(strip.text = ggtext::element_markdown(face = "bold"))\n<\/pre>\n\n\n
\"\"<\/a><\/figure>\n\n\n\n
We can say that the gpa<\/code><\/em> and usdtry<\/code><\/em> variables are in the same cluster and look very similar. I wonder if we will see it in our model as well.<\/p>\n\n\n
\n#Split into a train and test set\nsplits <- \n  df_tidy %>% \n  timetk::time_series_split(assess = "1 year", cumulative = TRUE)\n  \ntrain <- training(splits)\ntest <- testing(splits)\n\n#Modeling\ndf_rec <- \n  recipe(gpa ~ ., data = train) \n\n\ndf_spec <- \n  prophet_reg() %>%\n  set_engine(engine = "prophet")\n\n\nset.seed(12345)\ndf_wflow_fit <- \n  workflow() %>% \n  add_recipe(df_rec) %>% \n  add_model(df_spec) %>% \n  fit(train)\n\n#Calibration data\ndf_cal <- \n  modeltime_calibrate(df_wflow_fit, test)\n\n#Accuracy\ndf_cal %>% \n  modeltime_accuracy(metric_set = metric_set(rmse,rsq))\n\n# A tibble: 1 \u00d7 5\n#  .model_id .model_desc           .type  rmse   rsq\n#      <int> <chr>                 <chr> <dbl> <dbl>\n#1         1 PROPHET W\/ REGRESSORS Test   2.40 0.989\n<\/pre>\n\n\n
Based on the rsq<\/a><\/em> and rmse<\/a><\/em> values, our model looks very good. Now, we can calculate the variable importance values based on the model with the modeltStudio<\/a><\/em> package. This will pop up the interactive dashboard on your browser, so you will be able to select variables on the right bottom for partial dependence<\/a><\/em>.<\/p>\n\n\n
\n#VIP(variable importance)\n##Explainer object\nlibrary(DALEXtra)\nexplainer <- \n  DALEX::explain(\n    model = df_wflow_fit,\n    data  = train %>% select(-gpa),\n    y     = train$gpa,\n    label = "Prophet"\n  )\n\n##Model Studio\nset.seed(1983)\nmodelStudio::modelStudio(explainer, \n                         B = 100,\n                         viewer = "browser")\n<\/pre>\n\n\n
\"\"<\/a><\/figure>\n\n\n\n
Despite what economists say, the CDS(cds<\/code><\/em>) seems to have little effect on the fund based on the dashboard. Besides, it looks like the most dominant effect is the USD\/TRY(usdtry<\/code><\/em>) rates\u2019 and it has a strong linear relation to the GPA(gpa<\/code>) fund. This confirms the inference we found earlier with clustering.<\/p>\n\n
\r\n
To leave a comment<\/strong> for the author, please follow the link and comment on their blog:  DataGeeek<\/a><\/strong>.<\/div>\r\n
\r\nR-bloggers.com<\/a> offers daily e-mail updates<\/a><\/strong> about R<\/a> news and tutorials about learning R<\/a> and many other topics. Click here if you're looking to post or find an R\/data-science job<\/a>.\r\n\r\n
Want to share your content on R-bloggers? click here<\/a> if you have a blog, or  here<\/a> if you don't.\r\n<\/div>","protected":false},"excerpt":{"rendered":"
 The Eurobonds are so popular in Turkey nowadays. The reason for this is that the CDS of Turkey fell for quite a while since the new economic outlook began, according to the economists. We will look into the factors that shape this trend in this article. First, we build our …<\/div>\n
<\/div>\n
<\/div>\n","protected":false},"author":2348,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":[],"categories":[4],"tags":[],"aioseo_notices":[],"jetpack-related-posts":[],"amp_enabled":true,"_links":{"self":[{"href":"https:\/\/www.r-bloggers.com\/wp-json\/wp\/v2\/posts\/380223"}],"collection":[{"href":"https:\/\/www.r-bloggers.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.r-bloggers.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.r-bloggers.com\/wp-json\/wp\/v2\/users\/2348"}],"replies":[{"embeddable":true,"href":"https:\/\/www.r-bloggers.com\/wp-json\/wp\/v2\/comments?post=380223"}],"version-history":[{"count":8,"href":"https:\/\/www.r-bloggers.com\/wp-json\/wp\/v2\/posts\/380223\/revisions"}],"predecessor-version":[{"id":385380,"href":"https:\/\/www.r-bloggers.com\/wp-json\/wp\/v2\/posts\/380223\/revisions\/385380"}],"wp:attachment":[{"href":"https:\/\/www.r-bloggers.com\/wp-json\/wp\/v2\/media?parent=380223"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.r-bloggers.com\/wp-json\/wp\/v2\/categories?post=380223"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.r-bloggers.com\/wp-json\/wp\/v2\/tags?post=380223"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}