Research methodology

The importance to understand customer dropout and the diversity of employed algorithms requires an understanding of trends and existing problems to create a ground base of knowledge. For the development of the systematic literature review was adopted, the methodology applied by Kitchenham & Charters (2007) developed in three stages: Plan, Conduct and Report, as described in the following figure:

Research goals is to understand What is the current state of machine learning research studies to predict dropout in contractual settings? Based in this were identified five research questions to determine the main aspects related to the customer dropout with contractual settings:

RQ1. What studies have been published? 
RQ2. Which algorithms have been used to predict the dropout?
RQ3. What are the more relevant features related to predicting customer dropout? 
RQ4. When the dropout occurs? 
RQ5. What is the accuracy of the machine learning algorithms to predict dropout? 

Activity 6: State of the art

Consider the identification of type of machine learning algorithms according business area. This allows to identify which algorithms are used in each business area to address dropout.

Activity 7: Current situation

Methodology SLR

All searches where limited to the time interval between 2000 and 2020. Limited to articles in english.

Parsifal

("research papers about dropout with contractual settings" OR "membership") AND ("machine learning to predict dropout" OR "churn") AND ("investigations founds") AND ("not applicable for this research")

((“customer dropout”) OR (“customer churn”) AND “machine learning” AND (“contractual” OR “membership”))

Scopus

Search: 20_08_2020 https://www.scopus.com/results/results.uri?sort=plf-f&src=s&sid=841eb157c55d7eb4b10662211622e11a&sot=a&sdt=a&cluster=scosubtype%2c%22ar%22%2ct%2c%22cp%22%2ct%2bscolang%2c%22English%22%2ct&sl=149&s=ALL+%28+customer+AND+%28+%28dropout+%29+OR+%28+%22churn%22+%29+%29+AND+%22machine+learning%22+AND+%28+%22contractual%22+OR+%22membership%22+%29+%29+AND+PUBYEAR+%3e+1999+AND+PUBYEAR+%3c+2021&origin=cto&editSaveS=&txGid=751219e5fa081a46ec24053c168c70ad

Query with contractual

ALL ( customer AND ( ( dropout ) OR ( "churn" ) ) AND "machine learning" AND ( "contractual" OR "membership" ) ) AND PUBYEAR > 1999 AND PUBYEAR < 2021 AND ( LIMIT-TO ( DOCTYPE , "ar" ) OR LIMIT-TO ( DOCTYPE , "cp" ) ) AND ( LIMIT-TO ( LANGUAGE , "English" ) )
Results = 210

Query without contractual

ALL ( customer AND ( ( dropout ) OR ( "churn" ) ) AND "machine learning" ) AND PUBYEAR > 1999 AND PUBYEAR < 2021 AND ( LIMIT-TO ( DOCTYPE , "ar" ) OR LIMIT-TO ( DOCTYPE , "cp" ) ) AND ( LIMIT-TO ( LANGUAGE , "English" ) )

Results = 2000

Scopus Studies selection (bibtex)

IEEE Digital Library

Search: 21_08_2020 https://ieeexplore.ieee.org/search/searchresult.jsp?queryText=((%E2%80%9Ccustomer%20dropout%E2%80%9D)%20OR%20(%E2%80%9Ccustomer%20churn%E2%80%9D)%20AND%20%E2%80%9Cmachine%20learning%E2%80%9D%20AND%20(%E2%80%9Ccontractual%E2%80%9D%20OR%20%E2%80%9Cmembership%E2%80%9D))&highlight=true&returnFacets=ALL&returnType=SEARCH&matchPubs=true&ranges=2000_2020_Year

(((“customer") AND ("dropout" OR “churn”)) AND “machine learning” AND (“contractual” OR “membership”))

results:20

SpringerLink

Search: 21_08_2020 (("customer dropout"or "customer churn") and "machine learning" and (contractual* or membership)) https://link.springer.com/search?query=%28%28%22customer+dropout%22or+%22customer+churn%22%29+and+%22machine+learning%22+and+%28contract*+or+membership%29%29+&date-facet-mode=between&facet-start-year=2000&previous-start-year=1997&facet-end-year=2021&previous-end-year=2021

results:79

Science@Direct

Search: 21_08_2020 (((“customer") AND ("dropout" OR “churn”)) AND “machine learning” AND (“contractual” OR “membership”))

https://www.sciencedirect.com/search?qs=%28%28%28%E2%80%9Ccustomer%22%29%20AND%20%20%28%22dropout%22%20OR%20%E2%80%9Cchurn%E2%80%9D%29%29%20AND%20%E2%80%9Cmachine%20learning%E2%80%9D%20AND%20%28%E2%80%9Ccontractual%E2%80%9D%20OR%20%E2%80%9Cmembership%E2%80%9D%29%29&date=2000-2020&show=100&offset=100

Results:126

ISI Web of Science

Search: 21_08_2020 (("customer dropout"or "customer churn") and "machine learning" and (contractual or membership)) https://apps.webofknowledge.com/Search.do?product=UA&SID=E5YD2nJO9BseQ2L2G2V&search_mode=GeneralSearch&prID=d77c3de4-5ff9-4e31-a913-3dc06f30f206

TS=(customer AND ( dropout OR churn) AND "machine learning" ) Timespan: 2000-2020. Databases: WOS, CCC, DIIDW, KJD, MEDLINE, RSCI, SCIELO. Search language=Auto results:110

TS=(customer AND ( dropout OR churn) AND "machine learning" AND (contractual or membership)) Databases= WOS, CCC, DIIDW, KJD, MEDLINE, RSCI, SCIELO Timespan=2000-2020 Search language=Auto
results:6

ACM Digital Library

Search: 20_08_2020 https://dl.acm.org/action/doSearch?fillQuickSearch=false&expand=dl&AfterYear=2000&BeforeYear=2018&AllField=%28%E2%80%9Ccustomer+dropout%E2%80%9D%29+OR+%28%E2%80%9Ccustomer+churn%E2%80%9D%29+AND+%E2%80%9Cmachine+learning%E2%80%9D+AND+%28%E2%80%9Ccontractual%E2%80%9D+OR+%E2%80%9Cmembership%E2%80%9D%29

Full Query Sintax:
"query": { (“customer dropout”) OR (“customer churn”) AND “machine learning” AND (“contractual” OR “membership”) } 
"filter": { Publication Date: (01/01/2000 TO 12/31/2018), ACM Content: DL, NOT VirtualContent: true }

results: 8

Full Query Sintax: "query": { AllField:((“customer dropout”) OR (“customer churn”) AND “machine learning”) } "filter": { Publication Date: (01/01/2000 TO 12/31/2020), ACM Content: DL, NOT VirtualContent: true } results: 86

Data Analysis

Articles identified in the initial dataset

Source	Articles
Scopus	210
IEEE	20
Springer	79
Science Direct	126
WoS	6
ACM	8
Total	449

Science Direct incomplete items where 8, Scopus 11 and IEEE 1. Remaining 429. 16 duplicates removed 424. ASReview (van de Schoot et al., 2020) filtering process:

1. Selected randomly some articles and are identified as relevant and irrelevant, at least 5 each. ASReview suggests when to stop;
2. ASReview orders the publications in such a way that you see the most relevant publications first
3. Stopping criterium could be stopping after n presented abstracts were labeled irrelevant, or if your time is up. You can use the chart in the statistics panel to follow your progress"

Data processing is available in a R Script, quantitative analysis and some text mining in the selected articles