A Survey on Churn Analysis in Various Business Domains

Probably not a substantial claim: may not pass S3 (load-bearing).

Burez et al. utilized a method called the CUBE method to improve the churn prediction performance.

“Burez, Jonathan, and Dirk Van den Poel.(2009) utilized a method called the CUBE method to improve the churn prediction performance.[95].”

• cites [95] sourceCould not verify

  The paper’s introduction (p. 1) and evaluation (pp. 5–7) concern time-series forecasting and anomaly detection at Uber. Searches for Burez, Van den Poel, CUBE, and churn prediction found no passage addressing the attributed claim. (LLM, conf=0.76)

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Contractual churn is frequently found in a flat-rate service such as music and movie streaming services.

“In addition, contractual churn is frequently found in a flat-rate service such as music and movie streaming services.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

These days majority of log-based churn prediction problems use the probabilistic method to determine whether customers are churning or not and to give customers incentive to reuse their service.

“Therefore, these days majority of the log-based churn prediction problems use the probabilistic method to determine whether customers are churning or not and to give customers incentive to reuse their service.”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are 12 business fields that performed churn prediction.

“There are 12 business fields that performed churn prediction.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The time window method is frequently used when analyzing activity logs in a non-contractual situation.

“The time window method is frequently used when analyzing activity logs these days in a non-contractual situation.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The criteria for setting the time window are different for each service feature.

“The criteria for setting the time window are different for each service feature.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Modern churn prediction problems are mainly analyzed using log data.

“Modern churn prediction problems are mainly analyzed using log data.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The telecommunications industry accounts for the majority of previous studies on churn.

“The telecommunications industry accounts for the majority of previous studies on churn.”

Classified as probably not a substantial claim.

Churn has been defined in various ways in multiple industries.

“Churn has been defined in various ways in multiple industries.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The criteria for 'inactivity' and 'prolonged' are different according to each research field.

“However, the criteria for 'inactivity'and 'prolonged'are different according to each research field.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

In most churn prediction problems, the churn period is defined as a section that can restore customers' trust.

“Therefore, in most churn prediction problems, the churn period is defined as a section that can restore customers' trust.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Not all logs are helpful in estimating churn.

“However, not all logs are helpful in estimating churn.”

Probably not a substantial claim: may not pass S3 (load-bearing).

There were some studies which approached customer churn from a psychological perspective.

“There were some studies which approached customer churn from a psychological perspective.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The efficiency of a retained customer value is usually calculated with the CLV.

“The efficiency of a retained customer value is usually calculated with the CLV.”

May combine two claims.

Research on churn prediction was also conducted in online dating, online commerce, Q&A services, and social network-based services.

“Research on churn prediction was also conducted in online dating, online commerce, Q&A services, and social network-based services [46]”

• cites [46] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Logs remain in the majority of services.

“Logs remain in the majority of services.”

May combine two claims.

Studies on churn prediction were conducted in the human resources and energy fields although they were minority.

“Studies on churn prediction were conducted in the human resources and energy fields although they were minority.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

The newspaper subscription and music streaming service offer fixed-rate services, and customer churn is consistent with the contract renewal period.

“The newspaper subscription and mu-sic streaming service offer fixed-rate services, and customer churn is consistent with the contract renewal period.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Because the Internet service goes into an inactive state as customers wish, contract renewal takes place nearly real-time.

“On the other hand, because the Internet service goes into an inactive state as customers wish, contract renewal takes place nearlyreal-time.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Later on, studies on churn have been actively conducted in the gaming field as in the telecommunication field.

“Later on, studies on churn have been actively conducted in the gaming field as in the telecommunication field.”

May combine two claims.

Studies have emerged in recent years to develop other features that are as important as the last log the customer left before churning, or to discover potential churners by reprocessing the time series features.

“Therefore, studies have emerged in recent years to develop other features that are as important as the last log the customer left before churning, or to discover potential churners by reprocessing the time series features.”

Probably not a substantial claim: may not pass S3 (load-bearing).

RFM features are used to extract customers who carry business values.

“These features are used to extract customers who carry business values.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Gaming services have a fast cycle of customer inflow and churn because of mass competition.

“These services have a fast cycle of customer inflow and churn because of mass competition.”

May combine two claims.

There are multiple studies on the method of calculating the CLV and that apply the CLV to churn models.

“There are multiple studies on the method of calculating the CLV and that apply the CLV to churn models.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The value of m may be different for each customer segment.

“Here, the value of m may be different for each customer segment.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Churn is generally related to customers' last time activity.

“Churn is generally related to customers' last time activity.”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are 2 statistics churn prediction models in the finance business field.

“Finance 10 | 2 | | 1”

Probably not a substantial claim: may not pass S3 (load-bearing).

There is 1 deep learning churn prediction model in the telecommunications business field.

“Telecommunications 26 | 3 | 3 | 1”

Probably not a substantial claim: may not pass S3 (load-bearing).

According to the research of this paper, many articles use AUC as a performance measurement method aside from standard precision.

“According to the research of this paper, many articles use AUC as a performance measurement method aside from standard precision.”

Probably not a substantial claim: may not pass S3 (load-bearing).

There is 1 traditional machine learning churn prediction model in the internet service business field.

“Internet Service 1 | 1 | | 1”

Probably not a substantial claim: may not pass S3 (load-bearing).

There is 1 traditional machine learning churn prediction model in the psychology business field.

“Psychology 1 | | | |”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are 2 traditional machine learning churn prediction models in the retails business field.

“Retails 2 | | | |”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are 10 traditional machine learning churn prediction models in the finance business field.

“Finance 10 | 2 | | 1”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are 2 deep learning churn prediction models in the game business field.

“Game 16 | 7 | | 2”

Probably not a substantial claim: may not pass S3 (load-bearing).

There is 1 traditional machine learning churn prediction model in the marketing business field.

“Marketing 1 | | | |”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are 4 traditional machine learning churn prediction models in the newspapers business field.

“Newspapers 4 | | | |”

Probably not a substantial claim: may not pass S3 (load-bearing).

There is 1 traditional machine learning churn prediction model in the internet commerce business field.

“Internet Commerce 1 | | | |”

May combine two claims.

As for the financial and insurance sectors, since the log data is relatively small and the information obtained from customers does not change to a great degree, there are many statistical approaches using traditional machine learning models or survival analysis.

“As for the financial and insurance sectors, since the log data is relatively small and the information obtained from customers does not change to a great degree, there are many statistical approaches using traditional machine learning models or survival analysis.”

Probably not a substantial claim: may not pass S3 (load-bearing).

There is 1 statistics churn prediction model in the internet service business field.

“Internet Service 1 | 1 | | 1”

Probably not a substantial claim: may not pass S3 (load-bearing).

There is 1 traditional machine learning churn prediction model in the commerce business field.

“Commerce 1 | | | |”

Probably not a substantial claim: may not pass S3 (load-bearing).

There is 1 traditional machine learning churn prediction model in the music streaming service business field.

“Music Streaming Service 1 | | | |”

Probably not a substantial claim: may not pass S3 (load-bearing).

There is 1 deep learning churn prediction model in the insurance business field.

“Insurance 3 | | | 1”

Probably not a substantial claim: may not pass S3 (load-bearing).

There is 1 deep learning churn prediction model in the internet service business field.

“Internet Service 1 | 1 | | 1”

Probably not a substantial claim: may not pass S3 (load-bearing).

There is 1 traditional machine learning churn prediction model in the energy business field.

“Energy 1 | | | |”

Probably not a substantial claim: may not pass S3 (load-bearing).

There is 1 deep learning churn prediction model in the finance business field.

“Finance 10 | 2 | | 1”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are 3 statistics churn prediction models in the telecommunications business field.

“Telecommunications 26 | 3 | 3 | 1”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are 3 graph theory churn prediction models in the telecommunications business field.

“Telecommunications 26 | 3 | 3 | 1”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are 26 traditional machine learning churn prediction models in the telecommunications business field.

“Telecommunications 26 | 3 | 3 | 1”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are 16 traditional machine learning churn prediction models in the game business field.

“Game 16 | 7 | | 2”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are 7 statistics churn prediction models in the game business field.

“Game 16 | 7 | | 2”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are 3 traditional machine learning churn prediction models in the insurance business field.

“Insurance 3 | | | 1”

Probably not a substantial claim: may not pass S3 (load-bearing).

There is 1 traditional machine learning churn prediction model in the human resources business field.

“Human Resources 1 | | | |”

Probably not a substantial claim: may not pass S3 (load-bearing).

Non-contractual customer churns occur due to low switching cost for changing the service.

“These non-contractual customer churns occur due to low switching cost for changing the service [31].”

• cites [31] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

Lee et al. (2018) defined the period during which 75% of customers were continuously unconnected as a churn section for PC game services.

“Lee, Eunjo, et al. (2018) defined the period during which 75% of customers were continuously unconnected as a churn section by taking into consideration the characteristics of PC game services [17].”

• cites [17] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

According to Lee et al. (2018), the period during which 75% or more customers are continuously unconnected is 14 weeks, and if a customer is unconnected for 14 weeks, he/she is regarded as churn.

“According to Fig. 2, the period during which customers were unconnected more than 75% was 14 According to the figure, the period during which 75% or more customers are continuously unconnected is 14 weeks.Therefore, if a customer is unconnected for 14 weeks, he/she is regarded as churn and this period is a time window section.”

• cites [17] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Through such observation, the expected net business value can be increased by predicting customer churn rates and providing possible churn customers with incentives to retain them from leaving.

“And through such observation, the expected net business value can be increased by predicting customer churn rates and providing possible churn customers with incentives to retain them from leaving [16] [23].”

• cites [16]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [23] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

If the time period during which a customer completely churns is selected as a time window, the period for churn definition exponentially increases and it does not provide any gain in terms of business.

“If the time period during which a customer completely churns is selected as a time window, the period for churn definition exponentially increases and it does not provide any gain in terms of business as changing the will of the customers who want to churn is deemed impossible [17].”

• cites [17] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

There are two customer churn types, contractual churn and non-contractual churn, and three churn observation criteria: monthly, daily, and binary.

“As described above, there are two customer churn types, which are contractual churn and non-contractual churn.Additionally, there are three churn observation criteria as follows: monthly, daily, and binary.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Yang et al. (2019) found that more than 95% of customers did not return when they were absent for 3 consecutive days in mobile games, and set 3 days as the time window churn period.

“Yang, Wanshan, et al. (2019) analyzed log data to define the churn period of mobile games, and the analysis result showed that more than 95% of customers did not return when they were absent for 3 consecutive days.They set 3 days as the time window churn period [43].”

• cites [43]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

The reason for having multiple ways of defining customer churn is to periodically monitor the customers' status changes.

“The reason for having multiple ways of defining customer churn is to periodically monitor the customers' status changes.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Internet services interpret log-in as retention and unconnected access for a certain period as churn.

“Therefore, the Internet service interprets the log-in as prolonged, that is, the retention of the service, and interprets unconnected access for a certain period of time as churn [17].”

• cites [17] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

Churn criteria can be divided into contractual churn and non-contractual churn.

“Thus, we can divide the criteria of churn into contractual churn and non-contractual churn.”

Probably not a substantial claim: may not pass S2 (checkable).

It is necessary to extract important information using an appropriate feature engineering technique for churn prediction.

“Therefore, it is necessary to extract important information using an appropriate feature engineering technique for churn prediction.”

• AssumptionRests on an assumption: Service indicators are generally sparse and have many outliers, and the data distribution is skewed to one side.

Probably not a substantial claim: may not pass S3 (load-bearing).

There are four domains as a method of building a churn prediction model: traditional machine learning, statistics, graph theory, and deep learning.

“There are four domains as a method of building a churn prediction model.They are traditional machine learning, statistics, graph theory, and deep learning.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Binary encoding produces distance between categories.

“likewise binary encoding produces distance between categories.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Zhang et al. used latent feature modification for churn prediction and classified features into a memorization feature and a generalization feature depending on data attributes.

“Zhang, Rong, et al. (2017) also used latent feature modification for churn prediction.They classified features into a memorization feature and a generalization feature depending on data attributes [11].”

• cites [11] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

To calculate CLV within the contractual settings, a shifted-Beta-Geometric (sBG) model is used.

“To calculate CLV within the contractual settings, a shifted-Beta-Geometric (sBG) model is used.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Without normalization, the machine learning model generally have to select a very small learning rate when searching for the optimum, resulting in a long training time.

“Without normalization, the machine learning model generally have to select a very small learning rate when searching for the optimum, resulting in a long training time.”

• AssumptionRests on an assumption: The distribution of service data is generally skewed to one side.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Bucketing can be used for both categorical and continuous features.

“Bucketing (also called Discrete binning or Data binning) can be used for both categorical and continuous features.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Filling missing data improved performance instead of dropping them.

“Nimmagadda, Sravya, Akshay Subramaniam, and Man Long Wong.(2017) argued that filling the missing data improved performance instead of dropping them [58].”

• cites [58] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The generalization feature exhibits the characteristics of the churn/no-churn group with only short-term attributes.

“As opposed to the memorization feature, the generalization feature exhibits the characteristics of the churn/no-churn group with only shortterm attributes.”

Classified as probably not a substantial claim.

If a customer churns from a service, the company will have to recruit another customer by spending the CAC to maintain the service.

“If a customer churns from a service, the company will have to recruit another customer by spending the CAC to maintain the service.”

• AssumptionRests on an assumption: The CAC cost is the minimum value to operate service makes from a customer with Return on Investing (ROI).

Probably not a substantial claim: may not pass S3 (load-bearing).

The latent features generated during the autoencoder process compress high-dimensional sparse data into low-dimensional dense data.

“The latent features generated during this process compresses high-dimensional sparse data into low-dimensional dense data.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing). May combine two claims.

Data imputation techniques may help to solve data imbalances and make the best of the information used in the model.

“Data imputation techniques may help to solve data imbalances and make the best of the information used in the model.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Lee et al. claimed that in the customer churn prediction competition hosted by the authors, the winning team showed significantly better performance than other teams by utilizing the autoencoder to compress the sparse data.

“Lee, Eunjo, et al. (2018) claimed that in the customer churn prediction competition hosted by the authors, the winning team showed significantly better performance than other teams by utilizing the autoencoder to compress the sparse data [10].”

• cites [10] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

Nimmagadda et al. used log-normalization or quantile normalization to build a prediction model other than 0-1 normalization.

“Nimmagadda, Sravya, Akshay Subramaniam, and Man Long Wong.(2017) used techniques such as log-normalization or quantile normalization to build a prediction model other than 0-1 normalization [58].”

• cites [58] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

Probability models have been traditionally used for customer-base analysis.

“In particular, probability models have been traditionally used for customer-base analysis [149].”

• cites [149] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

If a single service is run for a long time, the service competition intensifies and the Customer Acquisition Cost (CAC) tends to increase.

“However, if a single service is run for a long time, the service competition intensifies and the Customer Acquisition Cost (CAC) tends to increase [16]”

• cites [16]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

The memorization feature data slowly reveal the characteristics of the churn/no-churn group usually over a period of time.

“These types of data slowly reveal the characteristics of the churn/no-churn group usually over a period of time.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Numerical encoding breaks the nominal nature inherent in categorical data and results in encoding with linearity between categories.

“Numerical encoding breaks the nominal nature inherent in categorical data and results in encoding with linearity between categories”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

The boundary is blurring between the above four disciplines in recent times.

“The boundary is blurring between the above four disciplines in recent times.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

One-hot encoding on categorical data produces an orthogonal feature space for each category.

“One-hot encoding on categorical data in this way produces an orthogonal feature space for each category.”

May combine two claims.

Normalized features can achieve rapid model training while using relatively large learning rates compared to the initial data.

“Normalized features can achieve rapid model training while using relatively large learning rates compared to the initial data.”

May combine two claims.

If a prediction model is trained on imbalanced data and always outputs only retained customers, it would maintain high accuracy but not be effective in identifying characteristics of churned customers.

“If a prediction model is trained using this data and it always outputs only the results that indicate customers being retained, the model would maintain a 96 percent accuracy.In this case, although the accuracy of the model may be said to be high, the model would not be effective in identifying characteristics of churned customers.”

• AssumptionRests on an assumption: In a stable service, the number of churned customers generally account for a small proportion compared to the number of retained customers.

Probably not a substantial claim: may not pass S3 (load-bearing).

The churn feature that combines the memorization feature and the generalization feature predicts churn by combining deep learning models and traditional machine learning techniques.

“The churn feature that combines the memorization feature and the generalization feature predicts churn by combining deep learning models and traditional machine learning techniques.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The generalization feature data predicts results with shallow machine learning models such as logistic linear regression.

“The generalization feature data predicts results with shallow machine learning models such as logistic linear regression.”

May combine two claims.

Sifa et al. improved performance by using a semi-supervised learning technique since there was no sufficient data to solve prediction problems.

“Sifa, Rafet, et al. (2015) improved performance by using a semi-supervised learning technique since there was no sufficient data to solve prediction problems [52].”

• cites [52] sourceNot checked — no full text of the source

Classified as probably not a substantial claim.

These changes have the side effect of the model learning unintentional continuity.

“These changes have the side effect of the model learning unintentional continuity.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Churn prediction tends to be more accurate when processing raw logs.

“However, churn prediction tends to be more accurate when processing raw logs.”

• AssumptionRests on an assumption: Service indicators are generally sparse and have many outliers, and the data distribution is skewed to one side.

Probably not a substantial claim: may not pass S4 (based).

In general, when customer behavior is stored as log data in gaming, Internet service, and telecommunication service fields, the customer data is often found in the form of high-dimensional sparse data.

“In general, when the customer behavior is stored as log data in gaming, Internet service, and telecommunication service fields, the customer data is often found in the form of highdimensional sparse data.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

For features that are too sparse to be used in the model due to large variance because the feature value category is too wide, bucketing makes these features into categorical features.

“For features that are too sparse to be used in the model due to the large variance because the feature value category is too wide, bucketing is a technique that makes these features into categorical features.”

Probably not a substantial claim: may not pass S3 (load-bearing).

According to Amin et al., the Mega-Trend Diffusion Function (MTDF) method provided the highest accuracy when compared with SMOTE, ADASYN, MWMOTE, ICOTE and TRkNN.

“According to the study, the Mega-Trend Diffusion Function (MTDF) method provided the highest accuracy when compared with other techniques such as Synthetic Minority Oversampling Technique (SMOTE), Adaptive Synthetic Sampling Approach (ADASYN), Majority Weighted Minority Oversampling Technique (MWMOTE), Immune Centroids Oversampling Technique (ICOTE) and couples Top-N Reverse K-Nearest Neighbor (TR”

• cites [144] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

The value of a churn prevention model can be calculated by multiplying the loss cost of one customer by the number of people who are prevented from churn with a churn prediction algorithm.

“In this way, we can calculate the value of a churn prevention model by multiplying the loss cost of one customer by the number of people who are prevented from churn with a churn prediction algorithm.”

• AssumptionRests on an assumption: Customer churn behavior is quantitative.
• AssumptionRests on an assumption: It is difficult to directly relate customer churn to a decrease in sales.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

With the emergence of deep learning techniques of machine learning, it has become possible to manage time-dependent high-dimensional sparse data.

“However, with the emergence of deep learning techniques of machine learning, it has become possible to manage timedependent high-dimensional sparse data.”

Probably not a substantial claim: may not pass S3 (load-bearing).

A deep learning algorithm enables generating latent features by compressing long-term features.

“Moreover, a deep learning algorithm enables generating latent features by compressing long-term features that is a new technique of feature embedding.”

May combine two claims.

A balanced dataset composed of churned and retained customers has higher noise tolerance than an imbalanced dataset, hence it is more likely to be able to obtain decision boundaries for minor groups.

“This is because a balanced dataset composed of churned and retained customers has higher noise tolerance than a imbalanced dataset, hence it is more likely to be able to obtain decision boundaries for minor groups, which in this case denotes the churned customer dataset [91].”

• cites [91] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Gui applied undersampling, oversampling, and SMOTE methods to an imbalanced dataset of telecommunication field and compared the performance of the derived churn prediction models.

“Gui, Chun.(2017) applied undersampling, oversampling, and SMOTE methods to an imbalanced dataset of telecommunication field and compared the performance of the derived churn prediction models [145].”

• cites [145] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

In the past, the customer behavior data was simplified through a bucketing process.

“Hence, in the past, the customer behavior data was simplified through a bucketing process.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Gui suggested that the SMOTE sampling technique provided the best prediction performance.

“In the study, the author suggested that the SMOTE sampling technique provided the best prediction performance.”

• cites [145] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

In statistics, probability models have been mainly used for conducting churn predictions.

“In statistics, probability models have been mainly used for conducting churn predictions.”

Probably not a substantial claim: may not pass S3 (load-bearing).

In a customer-base analysis, churn rate is applied to the survival time estimation when calculating the CLV.

“In a customer-base analysis, churn rate is applied to the survival time estimation when calculating the CLV.”

Probably not a substantial claim: may not pass S2 (checkable).

Churn studies on different fields are helpful in grasping the comprehensive view of churn and exploring various features to apply them to the churn models.

“Churn studies on different fields are undoubtedly helpful in grasping the comprehensive view of churn and exploring various features to apply them to the churn models.”

Probably not a substantial claim: may not pass S3 (load-bearing).

According to Goodfellow et al., deep learning is part of machine learning.

“According to Goodfellow, Ian, Yoshua Bengio, and Aaron Courville.(2016), deep learning is part of machine learning [64].”

• cites [64] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Deep learning techniques often used for predicting customer churn mainly involves training via sparse customer data that have been densified, or a fully-connected neural network produced from extracting the latent vector from the autoencoder and concatenating those features with static features.

“Deep learning techniques often used for predicting customer churn mainly involves training via sparse customer data that have been densified, or a fully-connected neural network produced from extracting the latent vector from the autoencoder and concatenating those features with static features.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

In the non-contractual settings, customer churn tends to have a characteristic of continuous probabilistic.

“In the non-contractual settings, customer churn tends to have a characteristic of continuous probabilistic.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Most churn prediction models are related to customer relation management.

“Most churn prediction models are related to customer relation management.”

• AssumptionRests on an assumption: Most churn prediction models are related to customer relation management.

Probably not a substantial claim: may not pass S4 (based).

The time window is used to select potential churning customers rather than measuring a customer's complete churn.

“Therefore, the time window is used to select potential churning customers rather than measuring a customer's complete churn.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The sBG model allows continuous interpretation of the conditions in which the customer retention is determined to be discrete-time contractual in a contractual service.

“Accordingly, the sBG model allows continuous interpretation of the conditions in which the customer retention is determined to be discrete-time contractual in a contractual service [150].”

• cites [150] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

There may be performance differences depending on whether the churn prediction model is robust against false positives or false negatives.

“For example, there may be performance differences depending on whether the churn prediction model is robust against false positives or false negatives.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

A ROC curve is a very robust measurement criterion that measures classifiers independent of class distribution and misclassification error cost.

“A ROC curve is a very robust measurement criterion that measures classifiers independent of class distribution and misclassification error cost.”

May combine two claims.

Researchers may be able to discern whether their constructed model is used widely through our study, but they would not be able to determine which model is suitable and has the best performance for their study.

“Accordingly, although researchers may be able to discern whether if their constructed model is used widely through our study, they would not be able to determine which model is suitable and has the best performance for their study.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

The statistical model used here is based on statistical inference and hypothesis testing, and survival analysis with hazard methods are used to build churn prediction models.

“The statistical model used here is based on statistical inference and hypothesis testing, and survival analysis with hazard methods are used to build churn prediction models.”

Probably not a substantial claim: may not pass S3 (load-bearing).

By combining the buyer behavior and survival distribution, the CLV can be calculated, referred to as the Pareto/NBD model.

“By combining the buyer behavior and survival distribution, the CLV can be calculated.and this method is simply referred to as the Pareto/NBD model.”

Classified as probably not a substantial claim.

The ROC curve is a performance measurement method that focuses on the prediction of churn customers.

“Therefore, it is a performance measurement method that focuses on the prediction of churn customers.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Recently, deep learning techniques have emerged as a method of predicting customer churn.

“Recently, deep learning techniques have emerged as a method of predicting customer churn.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Churn prediction analysis using deep learning algorithms has recently emerged.

“Churn prediction analysis using deep learning algorithms has recently emerged.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Once a graph is built, it searches churning customers through the graph correlation analysis.

“Once a graph is built, it searches churning customers through the graph correlation analysis.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The distribution of churn used gamma mixture of exponential, also known as the Pareto (of the second kind) distribution.

“Further, the distribution of churn used gamma mixture of exponential, which is also known as the Pareto (of the second kind) distribution.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

Apart from this, it seems to be different since the churn model that best interprets the relevant log data is applied.

“Apart from this, it seems to be different since the churn model that best interprets the relevant log data is applied.”

Probably not a substantial claim: may not pass S3 (load-bearing).

In the non-contractual settings, the repeat-buying behavior of customers have been previously expressed through negative binomial distributions (NBD).

“In the non-contractual settings, the repeat-buying behavior of customers have been previously expressed through negative binomial distributions (NBD) [151].”

• cites [151]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

Churn prediction models using deep learning is a new solution with a good structure for predicting modern churn datasets.

“Churn prediction models using deep learning is a new solution with a good structure for predicting modern churn datasets.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Traditional churn prediction algorithms including statistics methods are still actively used today.

“As mentioned earlier in the text, traditional churn prediction algorithms including statistics methods are still actively used today.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The graph theory identifies churn as a mathematical relationship.

“The graph theory identifies churn as a mathematical relationship.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

In the past, when predicting customer churn, researchers used survival analysis or time series analysis using statistics, graph theory, and traditional machine learning algorithms.

“In the past, when predicting customer churn, researchers used survival analysis or time series analysis using statistics, graph theory, and traditional machine learning algorithms.”

Probably not a substantial claim: may not pass S4 (based).

The reason why the preferred model for each business field is different is that the types and cycles of log data used for each business are different.

“The reason why the preferred model for each business field is different is that the types and cycles of log data used for each business are different.”

• AssumptionRests on an assumption: The types and cycles of log data used for each business are different.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Machine learning techniques also began to be utilized in customer churn for non-contractual settings.

“Machine learning techniques also began to be utilized in customer churn for non-contractual settings.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Deep learning also has many variances and tend to be classified separately from conventional machine learning.

“However, deep learning also has many variances and tend to be classified separately from conventional machine learning.”

Probably not a substantial claim: may not pass S4 (based).

Unlike other modeling techniques, the deep learning model is capable of converting high dimensional sparse log data into low dimensional dense features by embedding time series features.

“Unlike other modeling techniques, the deep learning model is capable of converting high dimensional sparse log data into low dimensional dense features by embedding time series features.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Deep learning is an intense extension of machine learning with neural network algorithms.

“Deep learning is an intense extension of machine learning with neural network algorithms.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Churn analysis is used in the fields of Internet services and games, insurance, and management.

“Churn analysis is used in the fields of Internet services and games, insurance, and management.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Loss costs for customer churn are calculated by CAC or CLV.

“Loss costs for customer churn are calculated by CAC or CLV.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Because deep learning's academic significance has recently grown, it has established itself as a single academic field.

“However, because its academic significance has recently grown, it has established itself as a single academic field.”

Classified as probably not a substantial claim. May combine two claims.

Deep learning algorithms outperform other algorithms likely due to large quantities of customer log data being collected via computers and the churn prediction model utilizing the entire set of this acquired data to make churn predictions.

“This is likely due to large quantities of customer log data being collected via computers and the churn pre-diction model utilizing the entire set of this acquired data to make churn predictions.”

Probably not a substantial claim: may not pass S4 (based).

The deep learning model can learn customer's behavioral patterns from vast amount of data by layer-wise stacked neurons structure.

“Also the deep learning model can learn customer's behavioral patterns from vast amount of data by layer-wise stacked neurons structure.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Some churn studies have often used a top 10% decline lift performance metric.

“On the other hand, some churn studies have often used a top 10% decline lift performance metric [13] [67].”

• cites [13]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [67]Not checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

The Pareto/NBD method has been actively used as a probability model for deriving the CLV until recently.

“The Pareto/NBD method has been actively used as a probability model for deriving the CLV until recently [19] [147] [148].”

• cites [19] sourceNot checked — no full text of the source
• cites [147] sourceNot checked — no full text of the source
• cites [148] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

In the non-contractual settings, customer churn is not easy to define and trace.

“In the non-contractual settings, customer churn is not easy to define and trace.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

Research on churn prediction usually begins to improve business outcomes.

“Research on churn prediction usually begins to improve business outcomes.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The beta-geometric/beta-binomial (BG/BB) model can be used; the beta-geometric model fits the retention rate and the beta-binomial distribution fits the consumer purchasing behavior.

“As another method, the beta-geometric/beta-binomial (BG/BB) model can be used.In this model, the beta-geometric model fits the retention rate and the beta-binomial distribution fits the consumer purchasing behavior [149].”

• cites [149] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Compared to statistical methods, machine learning techniques have robust non-linear relationships between features and can learn heterogeneous effect when given diverse features.

“Compared to statistical methods, machine learning techniques have robust non-linear relationships between features and can learn heterogeneous effect when given diverse features.”