Page 114 - 臺大管理論叢第32卷第1期
P. 114
Internet Celebrity Economy: Exploring the Value of Viewers’ Comment Features and Live Streamers’
Marketing Strategies in Forecasting Revenue
each category, representing the words in the text sample matching that category. Notably,
the classification system includes categories for a variety of emotional dimensions, making
it sensitive to the differences among discrete emotions.
4. Forecasting Model of Internet Streamer Revenues
Prior studies on viewers' gift-sending behavior in live streaming have ignored the
effects of streamer heterogeneity (e.g., physical attractiveness or personalization). These
studies lack data at the individual-level and treat all data alike, which create severe
challenges to the analysis of marketing data, making it hard to execute in a rigorous
manner that lends itself to replication by other researchers (Allenby et al., 2005). The
HB model helps solve the problem of insufficient sample information and provide
better forecasting accuracy (Lin, Chen, and Jen, 2007; Lynch, 2007). This study adopts
the HB model to study the effects of viewers' comments and streamers' behaviors on
viewers' gift-sending behavior in live streaming under the premise of considering
streamer heterogeneity. Specifically, we measure the revenue of streamers (y), donations
from viewers are included, on live-streaming platforms within the context of a normal
distribution. We also calculate streamer value E(y), which emphasizes potential revenue
streams as well. A standard HB model equation is as follows.
th
th
It is assumed that y is the value of the j gift sent to the i streamer. Then, we
ij
have
(1)
~ �� �� , � , ( 1 )
�
�� ~ �� �� , � , ( 1 )
�
�� � �� � ��
where are dummy variables denoting the vectors of the comment metrics (i.e., total
�� are dummy variables denoting the vectors of the comment metrics (i.e., total
where
where x are dummy variables denoting the vectors of the comment metrics (i.e., total
��
ij
number, positive comments, and negative comments), discrete emotional comments
number, positive comments, and negative comments), discrete emotional comments
number, positive comments, and negative comments), discrete emotional comments (i.e.,
(i.e., excited, amused, praising, complaining, disappointed, and ridiculing),
(i.e., excited, amused, praising, complaining, disappointed, and ridiculing),
excited, amused, praising, complaining, disappointed, and ridiculing), andstreamers’ be-
andstreamers’ behaviors (i.e., chatting with viewers, sharing food features, responding
andstreamers’ behaviors (i.e., chatting with viewers, sharing food features, responding
haviors (i.e., chatting with viewers, sharing food features, responding to questions from
to questions from viewers, showing the cooking process, and sharing tasting
to questions from viewers, showing the cooking process, and sharing tasting
th
viewers, showing the cooking process, and sharing tasting experiences) in the i live
th
experiences) in the i live streaming. is a × 1 ( = 14) matrix representing
th
experiences) in the i live streaming.
� is a × 1 ( = 14) matrix representing
streaming. β is a k × 1 (k = 14) matrix representing viewers’ gift-sending behavior struc-
�
i
viewers’ gift-sending behavior structure and is normally distributed.
viewers’ gift-sending behavior structure and is normally distributed.
ture and is normally distributed.
~ ( , ), (2)
(2)
� ~ ( , ), (2)
� � �
where is a × 1 ( = 5) matrix, which are dummy variables that denote the
� is a × 1 ( = 5) matrix, which are dummy variables that denote the
where
�
presence of streamer heterogeneity (i.e., gender, physical attractiveness, and hosting
presence of streamer heterogeneity (i.e., gender, physical attractiveness, and hosting
106
style). is represented by a × matrix that measures the relationship between
style). is represented by a × matrix that measures the relationship between
viewers’ gift-sending behavior structure and streamer heterogenity. is the positive
viewers’ gift-sending behavior structure and streamer heterogenity. is the positive
definite covariance matrix of × .
definite covariance matrix of × .
The model allows for heterogeneity in the mean vector and covariance matrix of
The model allows for heterogeneity in the mean vector and covariance matrix of
the normal distribution and thus reflects heterogeneous behaviors in terms of levels ( ) �
the normal distribution and thus reflects heterogeneous behaviors in terms of levels ( )
�
and variability ( ). The HB model with normal distribution assumes that the mean
�
and variability ( ). The HB model with normal distribution assumes that the mean
�
vector is normally distributed across viewers and employed an inverted gamma
vector
� is normally distributed across viewers and employed an inverted gamma
�
�
� , ).
distribution for , that is, �� ~ ( , ).
�
��
distribution for , that is, ~ ( � �
�
The core of the HB model accounts for the uncertainty of the sample (Chiang,
The core of the HB model accounts for the uncertainty of the sample (Chiang,
Chib, and Narasimhan, 1999). Accounting for uncertainty is critical whenever data
Chib, and Narasimhan, 1999). Accounting for uncertainty is critical whenever data
limitations exist that lead to imprecise inferences about any aspects of behavior
limitations exist that lead to imprecise inferences about any aspects of behavior
(Allenby et al., 2005). HB models are a combination of two things: (1) a model written
(Allenby et al., 2005). HB models are a combination of two things: (1) a model written
in a hierarchical form that is (2) estimated using Bayesian methods (Allenby et al.,
in a hierarchical form that is (2) estimated using Bayesian methods (Allenby et al.,
2005). We can analyze the marketing data using one model for within-unit analysis and
2005). We can analyze the marketing data using one model for within-unit analysis and
another model for across-unit analysis (Allenby et al., 2005). The within-unit model
another model for across-unit analysis (Allenby et al., 2005). The within-unit model
could be used to describe the viewers’ behaviors (e.g., viewers’ commenting or gift
could be used to describe the viewers’ behaviors (e.g., viewers’ commenting or gift
sending) over time, while the across-unit analysis could be used to describe the
sending) over time, while the across-unit analysis could be used to describe the
heterogeneity (e.g., the streamer’s outward beauty) of the units (Jen and Chen, 2007;
heterogeneity (e.g., the streamer’s outward beauty) of the units (Jen and Chen, 2007;
Rossi, Gilula, and Allenby, 2001). The sub-models are combined to form the
Rossi, Gilula, and Allenby, 2001). The sub-models are combined to form the
hierarchical model, and Bayes theorem is used to integrate the pieces and account for
hierarchical model, and Bayes theorem is used to integrate the pieces and account for
all the uncertainty that is present (Allenby et al., 2005). Furthermore, with the
all the uncertainty that is present (Allenby et al., 2005). Furthermore, with the
development of computational methods, Markov Chain Monte Carlo (MCMC) method
development of computational methods, Markov Chain Monte Carlo (MCMC) method
replaces the past complex analysis required to implement Bayes’ theorem (Haugh,
replaces the past complex analysis required to implement Bayes’ theorem (Haugh,
2017).
2017).
14 14