事實證明，性格偵測在許多領域是有用的。 五大性格特質模型和邁爾斯-布里格斯性格分類指標 是常見的性格選擇，但它們僅適用於某些行業，例如教育、商業。九型人格特徵可以提供更深入以及更易懂的個人資訊，並且更適合用在心理學和精神病學等領域。 本研究提出了基於九型人格特徵的人格分類，目的是提供自動性格偵測並了解每種性格類型用戶的在線互動的行為。 就結果來說，和詞袋功能相比，表情符號和語言特徵（例如語言查詢和單詞計數（LIWC）和詞性標籤（POS標籤））在跨源性格偵測中顯得更穩定和更泛化並且邏輯回歸證明在所有特徵和所有性格類型上都能正常工作。

Personality detection has been proved useful in many sectors. While Five Factor Model and Myers-Briggs Type Indication are common personality choice, they are only suitable for certain industries such as Education, Business. Enneagram personality could provide deeper and more understanding information of an individual and is more suitable for the fields like Psychology and Psychiatry. This work proposed the personality classification based on Enneagram personality with the aim of providing automatic personality detection framework and understanding the online interaction behaviors of the users in each personality type. As a result, compared to Bag-of-word features, emoticon and linguistic features such as Linguistic Inquiry and Word Counts (LIWC) and Part-of-speech tagging (POS Tag) appeared to be more robust and generalizable in cross-source personality detection while Logistic Regression proved to work generally well across all features and all personality types.

Contents
1 Introduction ........................................1
2 Enneagram Personality............................... 4
3 Related Work.........................................7
4 Methodology..........................................9
4.1Data Collection........................................9
4.2Data Pre-processing ..................................11
4.2.1Data Cleaning ......................................11
4.2.2Label Centers of Intelligence .................... 13
4.3Classification Model .................................14
4.4Features Extraction .................................15
5 Experiment & Results................................22
5.1Data Exploration..................................... 22
5.2Experiment & Framework ...............................28
5.2.1Multi Classification ...............................28
5.2.2Binary Classification ..............................30
5.2.3Binary to Multi Classification ...................40
5.3Feature Importance Analysis...........................41
6 Conclusions & Future Work...........................50
References...............................................52

[1] Basant Agarwal. Personality detection from text: a review.International Journal ofComputer System, 1(1):1–4, 2014.
[2] Morgan Alexander and Brent Schnipke. The enneagram: A primer for psychiatryresidents.American Journal of Psychiatry Residents’ Journal, 15:2–5, 03 2020.
[3] Danny Azucar, Davide Marengo, and Michele Settanni. Predicting the big 5 person-ality traits from digital footprints on social media: A meta-analysis.Personality andindividual differences, 124:150–159, 2018.[4] Marlene Cresci Cohen. Applications of the enneagram to psychological assessment.Enneagram. com, 2007.
[5] David Daniels, Terry Saracino, Meghan Fraley, Jennifer Christian, and Seth Pardo.Advancing ego development in adulthood through study of the enneagram system ofpersonality.Journal of adult development, 25(4):229–241, 2018.
[6] Sheng Fang, Catherine Achard, and S ́everine Dubuisson. Personality classification52
and behaviour interpretation: An approach based on feature categories. InProceed-ings of the 18th ACM International Conference on Multimodal Interaction, pages225–232, 2016.
[7] Golnoosh Farnadi, Susana Zoghbi, Marie-Francine Moens, and M. Cock. Recog-nising personality traits using facebook status updates.AAAI Workshop - TechnicalReport, pages 14–18, 01 2013.
[8] Matej Gjurkovi ́c and JanˇSnajder. Reddit: A gold mine for personality prediction.InProceedings of the Second Workshop on Computational Modeling of People’sOpinions, Personality, and Emotions in Social Media, pages 87–97, 2018.
[9] Jennifer Golbeck. Predicting personality from social media text.AIS Transactions onReplication Research, 2(1):2, 2016.
[10] Christopher L Heuertz.The sacred enneagram: Finding your unique path to spiritualgrowth. Zondervan, 2017.
[11] Francisco Iacobelli, Alastair J Gill, Scott Nowson, and Jon Oberlander. Large scalepersonality classification of bloggers. Ininternational conference on affective com-puting and intelligent interaction, pages 568–577. Springer, 2011.
[12] Mayuri Pundlik Kalghatgi, Manjula M. Ramannavar, and Dr. Nandini S. Sidnal. Aneural network approach to personality prediction based on the big-five model. 2015.
[13] Mi Jung Kang. Effects of enneagram program for self-esteem, interpersonal relation-ships, and gaf in psychiatric patients.Journal of Korean Academy of Psychiatric andMental Health Nursing, 24(2):116–126, 2015.53
[14] Michael C Komisin and Curry I Guinn. Identifying personality types using documentclassification methods. InTwenty-Fifth International FLAIRS Conference, 2012.
[15] Akshay Kulkarni and Adarsha Shivananda. Converting text to features. InNaturalLanguage Processing Recipes, pages 67–96. Springer, 2019.
[16] Weijian Li, Yuxiao Chen, Tianran Hu, and Jiebo Luo. Mining the relationship betweenemoji usage patterns and personality. InTwelfth International AAAI Conference onWeb and Social Media, 2018.
[17] Ana Carolina ES Lima and Leandro Nunes De Castro. A multi-label, semi-supervisedclassification approach applied to personality prediction in social media.Neural Net-works, 58:122–130, 2014.
[18] Fei Liu, Scott Nowson, and Julien Perez. A language-independent and compositionalmodel for personality trait recognition from short texts.ArXiv, abs/1610.04345, 2017.
[19] Edward Loper and Steven Bird. Nltk: the natural language toolkit.arXiv preprintcs/0205028, 2002.
[20] Anthony Ma. Neural networks in predicting myers brigg personality type from writingstyle. 2017.
[21] Davide Marengo, Fabrizia Giannotta, and Michele Settanni. Assessing personalityusing emoji: An exploratory study.Personality and Individual Differences, 112:74–78, 2017.
[22] Dejan Markovikj, Sonja Gievska, Michal Kosinski, and David J Stillwell. Mining facebook data for predictive personality modeling. InSeventh International AAAIConference on Weblogs and Social Media, 2013.
[23] Veronica Ong, Anneke DS Rahmanto, Derwin Suhartono, Aryo E Nugroho, Esther WAndangsari, Muhamad N Suprayogi, et al. Personality prediction based on twitterinformation in bahasa indonesia. In2017 Federated Conference on Computer Scienceand Information Systems (FedCSIS), pages 367–372. IEEE, 2017.
[24] Fabio Pianesi, Nadia Mana, Alessandro Cappelletti, Bruno Lepri, and Massimo Zan-canaro. Multimodal recognition of personality traits in social interactions. In Pro-ceedings of the 10th international conference on Multimodal interfaces, pages 53–60,2008.
[25] Barbara Plank and Dirk Hovy. Personality traits on twitter—or—how to get 1,500personality tests in a week. InProceedings of the 6th Workshop on ComputationalApproaches to Subjectivity, Sentiment and Social Media Analysis, pages 92–98, 2015.
[26] Bayu Yudha Pratama and Riyanarto Sarno. Personality classification based on twittertext using naive bayes, knn and svm. In2015 International Conference on Data andSoftware Engineering (ICoDSE), pages 170–174. IEEE, 2015.
[27] Daniel Preot ̧iuc-Pietro, Jordan Carpenter, and Lyle Ungar. Personality driven differ-ences in paraphrase preference. InProceedings of the Second Workshop on NLP andComputational Social Science, pages 17–26, 2017.
[28] Susan Reynolds, John K Waters, and Ronna Phifer-ritchie.The Everything Ennea-gram Book: Identify Your Type, Gain Insight into Your Personality and Find Successin Life, Love, and Business. Simon and Schuster, 2007.
[29] Mozhgan Saeidi, Saeid Komasi, Mohammad Amiri, Mona Azizi, and MostafaAlikhani. Is the enneagram personality system an effective approach in explainingdrug addiction?Middle East Journal of Rehabilitation and Health Studies, In Press,01 2020.
[30] Alireza Souri, Shafigheh Hosseinpour, and Amir Masoud Rahmani. Personality clas-sification based on profiles of social networks’users and the five-factor model ofpersonality.Human-centric Computing and Information Sciences, 8(1):24, 2018.
[31] Anna Sutton.Implicit and explicit personality in work settings: an application ofEnneagram theory. PhD thesis, University of Leeds, 2007.[32] Anna Sutton, Chris Allinson, and Helen Williams. Personality type and work-relatedoutcomes: An exploratory application of the enneagram model.European Manage-ment Journal, 31:234–249, 06 2013.
[33] Yla R Tausczik and James W Pennebaker. The psychological meaning of words: Liwcand computerized text analysis methods.Journal of language and social psychology,29(1):24–54, 2010.
[34] Sarah Theres V ̈olkel, Daniel Buschek, Jelena Pranjic, and Heinrich Hussmann. Un-derstanding emoji interpretation through user personality and message context. In Proceedings of the 21st International Conference on Human-Computer Interactionwith Mobile Devices and Services, pages 1–12, 2019.