隨著偶像經濟蓬勃發展，特典小卡（Pre-order Benefit, POB）在 C2C 次級市場中具備極高的流動性與投資價值。然而，小卡在預購階段缺乏歷史交易數據，且面臨強烈的資訊不對稱與週期性波動，導致傳統預測模型難以介入，產生資產評價的「冷啟動」難題。
為解決此痛點，本研究提出一套以「視覺特徵」為核心的廣義化 BG/BB (Beta-Geometric/Beta-Binomial) 模型。本研究將小卡的外觀屬性（如成員、眼神、嘴部表情、手部動作等）轉換為外部共變量整合至模型參數估計中，藉此量化視覺符號對市場買氣與流失風險的影響。同時，結合歷史特徵資料迴歸預測的動態期望價格，建構出冷啟動情境下的動態期望價值評估架構。
本研究以 Facebook「BABYMONSTER 台灣周邊交易社村落」社團為實證場域，蒐集共計 65 週的真實交易紀錄，並將樣本劃分為具歷史數據的「舊卡（訓練集）」與無歷史數據的「新卡（測試集）」。實證結果顯示：(1) 特徵整合的廣義化 BG/BB 模型能有效透過卡面視覺資訊，精準預測缺乏歷史數據之新卡的市場表現，成功克服冷啟動痛點，且新卡的預測命中率甚至優於舊卡；(2) 視覺特徵（如比愛心動作、微笑表情）對市場交易活躍度（購買率）具有顯著的正向影響；(3) 模型成功捕捉小卡次級市場特有的生命週期動態，包含新卡專屬的「新品蜜月期爆發」，以及舊卡面臨的「預算排擠效應」與高價值資產因惜售心理而產生的「收藏沉澱效應」（提早退出交易市場）。本研究不僅擴展了客戶基礎分析模型在非契約資產評價之應用，亦為經紀公司與投資型賣家提供具科學理性的視覺特徵配置與資產投資決策工具。

With the booming idol economy, Pre-order Benefit (POB) photocards have become highly liquid and valuable investment assets in the C2C secondary market. However, evaluating the value of these photocards during the pre-order phase is challenging due to the lack of historical transaction data, significant information asymmetry, and cyclical fluctuations. This creates a "Cold Start" problem that traditional forecasting models struggle to address. To solve this issue, this study proposes a feature-based generalized Beta-Geometric/Beta-Binomial (BG/BB) model. Visual features of photocards (e.g., members, eye directions, mouth expressions, and hand gestures) are integrated as external covariates into the model's parameter estimation to quantify the impact of visual symbols on market purchase rates and dropout risks. By combining this with dynamic expected prices predicted by historical feature regression, a dynamic expected value evaluation framework is established specifically for cold start scenarios.
Empirical data was collected from the Facebook trading group "BABYMONSTER Taiwan Merchandise Trading Village" over 65 weeks. The samples were divided into "old cards" (training set with historical data) and "new cards" (testing set without historical data). The empirical results demonstrate that: (1) the feature-integrated generalized BG/BB model effectively and accurately predicts the market performance of new cards lacking historical data by merely utilizing visual features, thereby successfully overcoming the cold start problem. The prediction accuracy for new cards even outperformed that of old cards; (2) specific visual features (e.g., heart gestures, smiling expressions) have a significant positive impact on market transaction activity; (3) the model successfully captures unique lifecycle dynamics in the photocard secondary market, including the "honeymoon period explosion" for new cards, and the early market exit of old cards due to the "budget crowding-out effect" and the "collection precipitation effect" driven by collectors' reluctance to sell high-value assets. This study not only extends the application of customer-base analysis models to non-contractual asset valuation but also provides entertainment agencies and investment-oriented sellers with a scientific decision-making tool for visual feature configuration and asset portfolio management.

Abe, M. (2009). "Counting your customers" one by one: A hierarchical Bayes extension to the Pareto/NBD model. Marketing Science, 28(3), 541–553.
Akerlof, G. A. (1978). The market for “lemons”: Quality uncertainty and the market mechanism. In Uncertainty in economics (pp. 235-251). Academic Press.
Amihud, Y., & Mendelson, H. (1986). Liquidity and stock returns. Financial Analysts Journal, 42(3), 43–48.
Ascarza, E., Fader, P. S., & Hardie, B. G. (2017). Marketing models for the customer-centric firm. In Handbook of marketing decision models (pp. 297-329). Cham: Springer International Publishing.
Breiman, L., Friedman, J., Olshen, R. A., & Stone, C. J. (2017). Classification and regression trees. Chapman and Hall/CRC.
Broyden, C. G. (1970). The convergence of a class of double-rank minimization algorithms 1. general considerations. IMA Journal of Applied Mathematics, 6(1), 76-90.
Chou, P., Chuang, H. H. C., Chou, Y. C., & Liang, T. P. (2022). Predictive analytics for customer repurchase: Interdisciplinary integration of buy till you die modeling and machine learning. European Journal of Operational Research, 296(2), 635-651.
Chung, J., Johar, G. V., Li, Y., Netzer, O., & Pearson, M. (2022). Mining consumer minds: Downstream consequences of host motivations for home-sharing platforms. Journal of Consumer Research, 48(5), 817-838.
Duffie, D., Gârleanu, N., & Pedersen, L. H. (2005). Over‐the‐counter markets. Econometrica, 73(6), 1815-1847.
Dutilleul, P. (1999). The MLE algorithm for the matrix normal distribution. Journal of Statistical Computation and Simulation, 64(2), 105–123.
Fader, P. S., Hardie, B. G., & Lee, K. L. (2005). "Counting your customers" the easy way: An alternative to the Pareto/NBD model. Marketing Science, 24(2), 275–284.
Fader, P. S., & Hardie, B. G. (2007). Incorporating time-invariant covariates into the Pareto/NBD and BG/NBD models. Retrieved July, 2, 2016.
Fader, P. S., & Hardie, B. G. (2009). Probability models for customer-base analysis. Journal of Interactive Marketing, 23(1), 61-69.
Fader, P. S., Hardie, B. G., & Shang, J. (2010). Customer-base analysis in a discrete-time noncontractual setting. Marketing Science, 29(6), 1086–1108.
Fan, Z. (2021). Improving Discrete Time BTYD Model with Covariates and Non-Parametric Priors (Doctoral dissertation, University of Pennsylvania).
Hartmann, T., & Goldhoorn, C. (2011). Horton and Wohl revisited: Exploring viewers' experience of parasocial interaction. Journal of Communication, 61(6), 1104–1121.
Hastie, T. (2009). The elements of statistical learning: data mining, inference, and prediction.
Hoerl, A. E., & Kennard, R. W. (1970). Ridge regression: Biased estimation for nonorthogonal problems. Technometrics, 12(1), 55–67.
Horton, D., & Richard Wohl, R. (1956). Mass communication and para-social interaction: Observations on intimacy at a distance. psychiatry, 19(3), 215–229.
Hwang, K., & Zhang, Q. (2018). Influence of parasocial relationship between digital celebrities and their followers on followers’ purchase and electronic word-of-mouth intentions, and persuasion knowledge. Computers in human behavior, 87, 155-173.
Maronna, R. A., Martin, R. D., Yohai, V. J., & Salibián-Barrera, M. (2019). Robust statistics: theory and methods (with R) (2nd ed.). John Wiley & Sons.
Platzer, M., & Reutterer, T. (2016). Ticking away the moments: Timing regularity helps to better predict customer activity. Marketing Science, 35(5), 779–799.
Rosen, S. (1974). Hedonic prices and implicit markets: product differentiation in pure competition. Journal of political economy, 82(1), 34–55.
Schein, A. I., Popescul, A., Ungar, L. H., & Pennock, D. M. (2002, August). Methods and metrics for cold-start recommendations. In Proceedings of the 25th annual international ACM SIGIR conference on Research and development in information retrieval (pp.253-260).
Schmittlein, D. C., Morrison, D. G., & Colombo, R. (1987). Counting your customers: Who-are they and what will they do next? Management Science, 33(1), 1–24.
Stever, G. S. (2011). Fan behavior and lifespan development theory: Explaining para-social and social attachment to celebrities. Journal of Adult Development, 18(1), 1-7.
Team, R. C. (2020). RA language and environment for statistical computing, R Foundation for Statistical. Computing.
Zonneveld, L., & Biggemann, S. (2014). Emotional connections to objects as shown through collecting behaviour: The role of Ardour. Australasian Marketing Journal, 22(4), 325-3.