FIKRIRAMASINGGIH.
Data Analytics Engineer
Data Engineer — Bank Mandiri via Lawencon Internasional
Data Analytics Engineer based in Jakarta, working across the full path from raw source systems to the query that answers a question. Currently a Data Engineer at Bank Mandiri, contracted through Lawencon Internasional, on Oracle stored procedures and ETL DataStage. Before that, a year as tier-3 data analyst at Telkomsel, and nearly three years at the consultancy PT Prima Integrasi — the last six months leading its data analytics team — implementing anti-money-laundering analytics on TigerGraph, building Oracle APEX applications, and assembling pipelines with Python, Spark and Informatica.
“Why make it complex when you can simplify it?”
- ROLE
- Data Analytics Engineer
- BASE
- Jakarta, Indonesia
- WORKING SINCE
- 2022
- PUBLISHED
- IEEE CITSM 2022
Education
Universitas Islam Negeri Syarif Hidayatullah Jakarta
Information Systems Department
Affiliation as registered with IEEE for the 2022 CITSM paper.
RESEARCH
IEEE CITSM 2022 · IEEE
Support Vector Machine and Lexicon based Sentiment Analysis on Kartu Prakerja (Indonesia Pre-Employment Cards Government Initiatives)
The study classifies public response to Kartu Prakerja, an Indonesian government programme providing assistance to people without work. Sentiment is derived from textual data to establish the public view of the policy. Weighting is based on matching text against a normalised sentiment lexicon, scaled to assign positive, neutral and negative classes; a Support Vector Machine is then trained and evaluated on the labelled set.
Kartu Prakerja reached 11.4 million recipients across 2020–2021 according to the Central Statistics Agency figures cited in the paper. Public opinion on the programme was divided, and that split is what the study set out to measure.
Authors
- Bayu Waspodo,
- Qurrotul Aini,
- Fikri Rama Singgih,
- Rinda Hesti Kusumaningtyas,
- Elvi Fetrina
Universitas Islam Negeri Syarif Hidayatullah Jakarta, Information Systems Department, Jakarta, Indonesia
- Venue
- 10th International Conference on Cyber and IT Service Management (CITSM)
- Held
- 20–21 September 2022 · Yogyakarta, Indonesia
- Publisher
- IEEE · pp. 01–06
Method
- 01
CRAWLING
Tweets collected as the raw corpus.
- 02
PREPROCESSING
Raw text normalised into comparable tokens.
CleaningLemmatizationStemmingTokenizingStopword removal - 03
LEXICON WEIGHTING
Tokens matched against a normalised sentiment lexicon and scaled into positive, neutral and negative classes.
- 04
SVM CLASSIFIER
Supervised classification, with the cost parameter selected using 10-fold cross validation.
- 05
SENTIMENT
Labelled output and visualisation of the distribution.
Classified sentiment
940 tweets
- POSITIVE
- 33035%
- NEUTRAL
- 30833%
- NEGATIVE
- 30232%
Reported performance
- Accuracy
- 98.75%
- Precision
- 0.98
- Recall
- 0.98
- F-measure
- 0.98
average
Figures are those reported by the authors on their own 940-tweet dataset. They describe performance on that dataset and are not a general claim about the methods.
KNOWLEDGE
Every node is something Fikri names publicly — on his GitHub profile, in the CITSM paper, or in a public repository. Nothing here is inferred from a job title.
Drag the graph to rotate it. Shift + scroll to zoom. Or use the list below — it does the same thing.
Select anything below, or a node in the graph, to see where it comes from.
PROJECTS
Each entry drives one of the structures in the scene behind this column. The form of each structure follows what the project actually is — a graph project is drawn as a graph, a pipeline as a flow.
Public repositories and published research. Fikri notes on his GitHub profile that some repositories stay private because the work is NDA-bound, so this is a sample of his output rather than a full record.
FIKRIRAMASINGGIH
Let’s connect
Data Analytics Engineer · Jakarta, Indonesia