Add call features

Snakefile

@@ -11,6 +11,16 @@ rule all:
                             sms_type = config["COM_SMS"]["SMS_TYPES"],
                             day_segment = config["COM_SMS"]["DAY_SEGMENTS"],
                             metric = config["COM_SMS"]["METRICS"]),
+        expand("data/processed/{pid}/com_call_{call_type}_{segment}_{metric}.csv",
+                            pid=config["PIDS"], 
+                            call_type = config["COM_CALL"]["CALL_TYPE_MISSED"],
+                            segment = config["COM_CALL"]["DAY_SEGMENTS"],
+                            metric = config["COM_CALL"]["METRICS_MISSED"]),
+        expand("data/processed/{pid}/com_call_{call_type}_{segment}_{metric}.csv",
+                            pid=config["PIDS"], 
+                            call_type = config["COM_CALL"]["CALL_TYPE_TAKEN"],
+                            segment = config["COM_CALL"]["DAY_SEGMENTS"],
+                            metric = config["COM_CALL"]["METRICS_TAKEN"]),
 
 # --- Packrat Rules --- #
 ## Taken from https://github.com/lachlandeer/snakemake-econ-r

config.yaml

@@ -1,5 +1,5 @@
 # Valid database table names
-SENSORS: [messages]
+SENSORS: [messages, calls]
 
 # Participants to include in the analysis
 # You must create a file for each participant
@@ -23,3 +23,12 @@ COM_SMS:
   SMS_TYPES : [received, sent]
   DAY_SEGMENTS: *day_segments
   METRICS: [count, distinctcontacts]
+
+# Communication call features config
+# Separate configurations for missed and taken calls
+COM_CALL:
+  CALL_TYPE_MISSED : [missed]
+  CALL_TYPE_TAKEN : [incoming, outgoing]
+  DAY_SEGMENTS: *day_segments
+  METRICS_MISSED: [count, distinctcontacts]
+  METRICS_TAKEN: [count, distinctcontacts, meanduration, sumduration, hubermduration, varqnduration, entropyduration]

packrat/packrat.lock

@@ -13,6 +13,11 @@ Source: CRAN
 Version: 1.0.0
 Hash: 6abedd7919c4457604c0aa44529a6683
 
+Package: DEoptimR
+Source: CRAN
+Version: 1.0-8
+Hash: adc74e88e85eabe6c7d73db6a86fe6cf
+
 Package: R6
 Source: CRAN
 Version: 2.4.0
@@ -125,6 +130,11 @@ Version: 0.3.0
 Hash: 30b58109e4d7c6184a9c2e32f9ae38c6
 Requires: rlang
 
+Package: entropy
+Source: CRAN
+Version: 1.2.1
+Hash: ccff926ff232f7c19b4c84bab3d3d6d3
+
 Package: evaluate
 Source: CRAN
 Version: 0.14
@@ -362,6 +372,12 @@ Hash: 1f3014c40b12e8af0abf39fd78080237
 Requires: base64enc, evaluate, htmltools, jsonlite, knitr, mime,
     stringr, tinytex, xfun, yaml
 
+Package: robustbase
+Source: CRAN
+Version: 0.93-5
+Hash: 7b6672bf2b47c35d02a5b273393e49f5
+Requires: DEoptimR
+
 Package: rstudioapi
 Source: CRAN
 Version: 0.10

rules/features.snakefile

@@ -9,3 +9,15 @@ rule communication_sms_metrics:
         "data/processed/{pid}/com_sms_{sms_type}_{day_segment}_{metric}.csv"
     script:
         "../src/features/communication_sms_metrics.R"
+
+rule communication_call_metrics:
+    input: 
+        "data/raw/{pid}/calls_with_datetime.csv"
+    params:
+        call_type = "{call_type}",
+        day_segment = "{day_segment}",
+        metric = "{metric}"
+    output:
+        "data/processed/{pid}/com_call_{call_type}_{day_segment}_{metric}.csv"
+    script:
+        "../src/features/communication_call_metrics.R"

src/features/communication_call_metrics.R

@@ -0,0 +1,30 @@
+source("packrat/init.R")
+
+library(dplyr)
+library(entropy)
+library(robustbase)
+
+calls <-  read.csv(snakemake@input[[1]])
+day_segment <- snakemake@params[["day_segment"]]
+metric <-  snakemake@params[["metric"]]
+type <- snakemake@params[["call_type"]]
+output_file <- snakemake@output[[1]]
+
+metrics <- calls %>% filter(call_type == ifelse(type == "incoming", "1", ifelse(type == "outgoing", "2", "3")))
+
+if(day_segment == "daily"){
+  metrics <- metrics %>% group_by(local_date)
+} else {
+  metrics <- metrics %>% filter(day_segment == local_day_segment) %>% group_by(local_date)
+}
+
+metrics <- switch(metric,
+       "count" = metrics %>% summarise(!!paste("com", "call", type, day_segment, metric, sep = "_") := n()),
+       "distinctcontacts" = metrics %>% summarise(!!paste("com", "call", type, day_segment, metric, sep = "_") := n_distinct(trace)),
+       "meanduration" = metrics %>% summarise(!!paste("com", "call", type, day_segment, metric, sep = "_") := mean(call_duration)),
+       "sumduration" = metrics %>% summarise(!!paste("com", "call", type, day_segment, metric, sep = "_") := sum(call_duration)),
+       "hubermduration" = metrics %>% summarise(!!paste("com", "call", type, day_segment, metric, sep = "_") := huberM(call_duration)$mu),
+       "varqnduration" = metrics %>% summarise(!!paste("com", "call", type, day_segment, metric, sep = "_") := Qn(call_duration)),
+       "entropyduration" = metrics %>% summarise(!!paste("com", "call", type, day_segment, metric, sep = "_") := entropy.MillerMadow(call_duration)))
+
+write.csv(na.omit(metrics), output_file, row.names = F)