Portfolio

Project description:

Skills & stack used:

• Technical documentation
• API documentation
• Programming languages: Python, Delphi
• Beginner-to-advanced Python programming concepts
• Python libraries
• Beginner-to-advanced Python use cases (including basic programming concepts, machine learning, deep learning, data analysis, data visualization, code testing, code profiling, GUI development, etc.)
• SEO

Link(s):

Project description:

• Build new graphical user interfaces (GUI) prototypes using Python4Delphi.
• Wrote longer and more advanced articles about various implementations and use cases of Python4Delphi compared to articles for pythongui.org.

Skills & stack used:

• GUI development
• Technical documentation
• Programming language: Python, Delphi
• Intermediate-to-advanced Python programming concepts
• Python libraries
• Intermediate-to-advanced Python use cases (including machine learning, deep learning, data analysis, data visualization, GUI development, etc.)
• SEO

Link(s):

Project description:

Skills & stack used:

• GUI development
• Create a presentation/course/demo video about programming
• Programming languages: Python, Delphi
• App development:
  • Scrapy + Delphi for web scraping (e.g., scrape all Google Scholar query search results)
  • Matplotlib + Delphi for data visualizations
  • fast.ai + Delphi for deep learning (implementation of ResNet-34 for image classification)
  • scikit-learn + Delphi for unsupervised machine learning (comparing performances of 10 different clustering algorithms)
  • NetworkX + Delphi for network visualizations
  • pandas + Delphi for tabular data analysis
• Technical documentation
• YouTube

Link(s):

Project description:

Skills & stack used:

• Technical documentation
• Programming Language: Python, Node.js, React.js, JavaScript, Go
• App development: Search engine, news aggregator, web scraper, emotion discovery from texts, profanity filter, etc.
• REST API
• GCP Products: Google Cloud IoT Core
• API testing with Postman
• SEO

Link(s):

Project description:

• Implement Latent Dirichlet allocation (LDA) to model the topics discussed by Indonesian religious leaders through their Instagram posts.
• The goal was to identify the 5-10 most dominant topics, ranked by their prevalence (frequency of discussion) and coherence (quality of topic modeling results).
• Reference for this project’s methodologies:
  • Jones, T. (2019). A coefficient of determination for probabilistic topic models. arXiv preprint arXiv:1911.11061.
  • Röder, M., Both, A., & Hinneburg, A. (2015, February). Exploring the space of topic coherence measures. In Proceedings of the eighth ACM international conference on Web search and data mining (pp. 399-408).
  • Rosner, F., Hinneburg, A., Röder, M., Nettling, M., & Both, A. (2014). Evaluating topic coherence measures. arXiv preprint arXiv:1403.6397.

Skills & stack used:

• NLP: Text preprocessing (R)
• NLP: Topic Modeling with Latent Dirichlet allocation/LDA (R)
• Data visualizations (Ms. Excel, R)

Project description:

Skills & stack used:

• Exploratory data analysis
• Data visualizations
• Ms. Excel

Project Description:

• This research aimed to determine the preferred food delivery service apps in Indonesia (GoFood vs GrabFood) during the early stages of the Covid-19 pandemic in 2020.
• The research utilized a dataset of 231,577 unique tweets for the “gofood” query search and 168,588 unique tweets for “grabfood query search.
• To uncover the preferred food delivery service apps in Indonesia during the 2020 pandemic, we employed sentiment analysis and topic modeling to uncover the reasons between users’ sentiments and preferences.
• The project proposed combining NLP (sentiment analysis and topic modeling) with a traditional qualitative-quantitative mixed-method approach to study the user intent behind user-generated content.
• Reference for this project’s methodologies:
  • Ray, A., & Bala, P. K. (2021). User generated content for exploring factors affecting intention to use travel and food delivery services. International Journal of Hospitality Management, 92, 102730.
  • Devereux, S., Béné, C., & Hoddinott, J. (2020). Conceptualising COVID-19’s impacts on household food security. Food Security, 12(4), 769-772.

Skills & Stack Used:

• Twitter scraping (Python)
• Text preprocessing (R)
• Text filtering & categorization (R)
• NLP: Sentiment Analysis with Dictionary-Based Approach (R)
• NLP: Topic Modeling with Latent Dirichlet allocation (LDA)
• Assigning the topic modeling results into several value & risk subcategories (human-assisted task)
• Data visualizations (Ms. Excel, R)

Project description:

• The objective of this project was to analyze the behavioral factors influencing food preferences in Indonesia and segment the Indonesian audience based on those preferences using social media data.
• The project has been published on Mendeley Data.
• Data was collected through an online platform by performing a query search on Facebook Audience Insights Interests.
• The query search keywords were based on the United Nations Food and Agriculture Organization (FAO) Food Balance Sheet (FBS), obtained from FAOStat in May 2020.
• Data collection took place between May 15 and July 2, 2020. The sample size consisted of 100-150 million users, which represents approximately 36.95% – 55.43% of Indonesia’s 2019 population. The sample was limited to Indonesia.

Skills & stack used:

• Data collection
• Facebook Audience Insights
• Statistics
• Data analysis

Link(s):

Project description:

• This research aimed to study value co-destruction in the Indonesian tourism site (case study: Komodo National Park) by performing a mixed-method study.
• The quantitative aspect involved employing Text Mining and Unsupervised Learning, while the qualitative aspect adopted service-dominant logic foundational premise modifications and additions from Vargo and Lusch (2008). These approaches were applied to the collected online reviews dataset.
• The dataset consisted of 41,914 records, comprising TripAdvisor and Google Review data about Komodo National Park, and all Indonesian tweets with the query search “pulau komodo” from January 2016 until December 2019.
• Sentiment Analysis with a Dictionary-Based Approach was performed to collect reviews with negative sentiments.
• Document Clustering (Agglomerative Hierarchical Clustering/AHC) was employed to analyze the negative review datasets and reveal the value co-destruction.
• The expected results were the clustering of documents into several groups, where reviews within each group had higher similarities compared to reviews in different groups.
• To gain insights about each cluster, the clusters were visualized as dendrograms and word clouds.
• Reference for this project’s methodologies:
  • Chakravarty, A., Liu, Y., & Mazumdar, T. (2010). The differential effects of online word-of-mouth and critics’ reviews on pre-release movie evaluation. Journal of interactive marketing, 24(3), 185-197.
  • Garg, N., & Gupta, R. K. (2016). Clustering techniques on text mining: a review. International Journal of Engineering Research, 5(4), 241-243.
  • Kassambara, A. (2017). Practical guide to cluster analysis in R: Unsupervised machine learning (Vol. 1). Sthda.
  • Vargo, S. L., & Lusch, R. F. (2004). Evolving to a new dominant logic for marketing. Journal of marketing, 68(1), 1-17.

Skills & stack used:

• Twitter scraping (Python)
• Text preprocessing (R)
• NLP: Sentiment Analysis with Dictionary-Based Approach (R)
• Text Mining & Unsupervised Learning: Cluster Analysis using Agglomerative Hierarchical Clustering/AHC (R)
• Data visualizations: Dendrogram, Wordcloud (R)

Project description:

• This research aimed to study the revealed vs unrevealed customers’ experiences on different review and social media platforms, focusing on the Indonesia tourism site, Komodo National Park.
• The quantitative aspect involved employing Sentiment Analysis and Text Mining, while the qualitative aspect adopted an explanatory sequential mixed-method research design from Creswell et al. (2003). These approaches were applied to the collected online reviews dataset.
• The dataset consisted of 41,914 records, comprising TripAdvisor and Google Review data about Komodo National Park, and all Indonesian tweets with the query search “pulau komodo” from January 2016 until December 2019.
• Sentiment Analysis with a Dictionary-Based Approach was performed separately on each dataset, based on the data source (TripAdvisor, Google Review, and Twitter).
• LDA Topic Modeling was used to unsupervisedly analyze the dataset and reveal the unrevealed/unstated customers’ experiences.
• The topic modeling results were then qualitatively classified into the following hypothetical categories: Tangible nouns, intangible nouns, material motives, social motives, noncommercial motives, and in-group identity signal.
• Furthermore, the change in hypothetical categories for the Twitter dataset was observed, and the longitudinal changes were presented with a plot.
• Reference for this project’s methodologies:
  • Creswell, J. W., Plano Clark, V. L., Gutmann, M. L., & Hanson, W. E. (2003). Advanced mixed methods research designs. Handbook of mixed methods in social and behavioral research, 209(240), 209-240.
  • Ravindran, S. K., and Garg, V. (2015). Mastering Social Media Mining with R. Packt Publishing.
  • Silge, J., & Robinson, D. (2017). Text mining with R: A tidy approach. “O’Reilly Media, Inc.”.

Skills & stack used:

• Twitter scraping (Python)
• Text preprocessing (R)
• NLP: Sentiment Analysis with Dictionary-Based Approach (R)
• NLP: Topic Modeling with Latent Dirichlet allocation/LDA (R)
• Mix-method research design
• Business knowledge
• Data visualizations (Ms. Excel, R)

Project description:

• This research aimed to improve products and services, as well as explore potential new products/services, by analyzing users’ feedback and questionnaires from Inspira Solution’s Workshop.
• The research also aimed to provide better, more relevant, and personalized content for valuable customers, regular customers, followers, and prospects.
• Text mining (TF-IDF) was implemented to extract the most important and relevant words or concerns expressed by our customers.
• The results were then visualized using word clouds, starting from the most general and important (N=100) to the more specific concerns (N=10), categorized into five categories: motivation, plans, type, preparations, and challenges of studying abroad.
• Reference for this project’s methodologies:
  • Feinerer, I. (2013). Introduction to the tm Package Text Mining in R. Accessible online via: cran.r-project.org.
  • Silge, J., & Robinson, D. (2017). Text mining with R: A tidy approach. “O’Reilly Media, Inc.”.

Skills & stack used:

• Indonesian text preprocessing (R)
• NLP/Text mining: TF-IDF (R)
• Data visualization: Wordcloud (R)
• Business knowledge

Project description:

• Scraped over 3.3 million rows of social media data (tweets from Twitter) related to the Indonesian Sharia Economy over the past 5 years (2014-2019).
• Web scraping was performed using Python to retrieve all tweets with the query search “halal” in the Indonesian Language from September 2014 to August 2019.
• All data was delivered in CSV files, with 60 files representing each month of tweets for my client.
• Reference for this project’s methodologies:
  • GetOldTweets-python: A project written in Python to retrieve old tweets, bypassing some limitations of the Twitter Official API. GitHub Repo [accessed Aug 2019].
  • snscrape: A social networking service scraper in Python. GitHub Repo [accessed Aug 2019].
  • Tweepy: Twitter for Python! GitHub Repo [accessed Aug 2019].

Skills & stack used:

Project description:

• Developed an ideal customer/buyer’s persona using an internal template provided by the Chief Marketing Officer (CMO), leveraging data from digital marketing platforms, customer feedback, and questionnaires.
• The primary objective of this research was to identify the most valuable customers, their segments, and profiles.
• Additionally, the project aimed to improve the quality of collected customer data.
• The insights gained from the buyer’s persona would inform future customer acquisition strategies.

Skills & stack used:

• Data analysis & data visualization (pivot table & pivot chart in Google Sheets)
• Business knowledge (buyer’s persona, customer acquisitions)
• Digital marketing (Facebook data driven marketing, e.g. Facebook Audience Insights, Facebook Lookalike Targetting, etc.)

Project description:

• Conducted research to estimate the market size for Inspira Solution, a critical factor in determining potential future completed transactions or sales. The findings served as a basis for setting realistic targets for total sales or completed transactions.
• The results of the market size estimation also played a crucial role in formulating future customer acquisition strategies and facilitating decision-making by C-level executives.

Skills & stack used:

• Digital marketing techniques (Facebook Audience, Facebook Audience Insights, Facebook Lookalike Audience 1% based on the most valuable customers, regular customers, leads, and social media/fan page followers)
• Business knowledge (market sizing methodologies: Total Addressable Market (TAM), Serviceable Available Market (SAM), Serviceable Obtainable Market (SOM), and Funneling)
• Data analysis
• Report writing

Project description:

• Conducted research to estimate the market size for IDPhotobook, a critical factor in determining potential future completed transactions or sales. The findings served as a basis for setting realistic targets for total sales or completed transactions.
• The results of the market size estimation also played a crucial role in formulating future customer acquisition strategies and facilitating decision-making by C-level executives.
• Compared to the previous result (portfolio 14), this market size estimation is more detailed/precise, as the data is richer due to the business is more established compared to the previous one (for example, to do a 1% Lookalike Audience, we can use the tens of thousands customer data from our internal database).

Skills & stack used:

• SQL query, to pull the data
• Digital marketing techniques (Facebook Audience, Facebook Audience Insights, Facebook Lookalike Audience 1% based on the most valuable customers, regular customers, leads, and social media/fan page followers)
• Business knowledge (market sizing methodologies: Total Addressable Market (TAM), Serviceable Available Market (SAM), Serviceable Obtainable Market (SOM), and Funneling)
• Data analysis
• Data visualization
• Report writing

Project description:

• This project was dedicated to the identification and calculation of the total number of repeat order customers, as well as the customer repeat order rate (including gaining insights into the current state of the business by examining the repeat order rate and identifying the most valuable customer segments).
• The data gathered and insights generated played a pivotal role in executing Facebook Lookalike targetting campaigns and formulating future marketing strategies.

Skills & stack used:

• Employed SQL database skills for effective data grouping and organization, enabling precise calculations and analysis (MySQL).
• Data Analysis and visualization techniques (Ms. Excel)
• End-to-end report writing, ensuring that the findings and recommendations were effectively communicated to C-level executives.
• Clear and impactful communication to C-Levels

Project description:

• Create an analytic dashboard prototype to be integrated into the internal real-time ERP Systems of IDPhotobook.
• Use Regression techniques to model not only overall sales performances but also the individual performance metrics of each Deal Maker within the organization (the users/C-levels can directly browse the contributions of each Deal Maker to the overall sales).
• The model is designed to monitor the performances over time (sales/month, week, or daily (browsed by date)).

Skills & stack used:

• Statistics & Data Analysis (Ms. Excel and Google Sheet)
• Machine Learning/Predictive Modeling: Linear Regression
• Relational database (MySQL)
• Data visualization (Ms. Excel, Amchart.js)