Analisis sentimen masyarakat pengguna media sosial terhadap program makan bergizi gratis menggunakan Algoritma IndoBERT

INDONESIA:

Program Makan Bergizi Gratis (MBG) merupakan salah satu program pemerintah yang menimbulkan berbagai tanggapan masyarakat di media sosial, baik berupa sentimen positif maupun negatif. Penelitian ini bertujuan untuk mengetahui kecenderungan sentimen masyarakat terhadap program Makan Bergizi Gratis (MBG) serta menganalisis kinerja algoritma IndoBERT dalam mengklasifikasikan sentimen teks berbahasa Indonesia. Penelitian dilakukan menggunakan data komentar media sosial dan diolah melalui tahapan preprocessing yang meliputi cleaning, normalisasi kata, tokenisasi, stopword removal, stemming, labelling, dan Random Oversampling. Hasil penelitian menunjukkan bahwa sentimen masyarakat terhadap program Makan Bergizi Gratis (MBG) cenderung didominasi oleh sentimen negatif dengan jumlah 2948 data, sedangkan sentimen positif berjumlah 1256 data. Selain itu, model IndoBERT mampu memberikan performa klasifikasi yang sangat baik dengan hasil terbaik diperoleh pada skenario 10 menggunakan rasio data 80:20, batch size 16, dan learning rate 0,00003 yang menghasilkan accuracy sebesar 95,85%, precision sebesar 94,42%, recall sebesar 97,46%, dan F1-score sebesar 95,91%.

ENGLISH:

The Free Nutritious Meal Program (MBG) is one of the government initiatives that has generated various public responses on social media, both in the form of positive and negative sentiments. This study aims to determine the tendency of public sentiment toward the Free Nutritious Meal Program (MBG) and to analyze the performance of the IndoBERT algorithm in classifying Indonesian-language text sentiment. The research was conducted using social media comment data processed through several preprocessing stages, including cleaning, word normalization, tokenization, stopword removal, stemming, labelling, and Random Oversampling. The results indicate that public sentiment toward the Free Nutritious Meal Program (MBG) tends to be dominated by negative sentiment, with a total of 2,948 data entries, while positive sentiment accounts for 1,256 data entries. Furthermore, the IndoBERT model demonstrated excellent classification performance, with the best results achieved in Scenario 10 using an 80:20 data split ratio, a batch size of 16, and a learning rate of 0.00003, yielding an accuracy of 95.85%, precision of 94.42%, recall of 97.46%, and an F1-score of 95.91%.

ARABIC:

ﻳﻌُﺪ ﺑﺮﻧﺎﻣﺞ »ﺍﻟﻮﺟﺒﺎﺕ ﺍﻟﻤﻐﺬﻳﺔ ﺍﻟﻤﺠﺎﻧﻴﺔ« (MBG) ﺃﺣﺪ ﺍﻟﺒﺮﺍﻣﺞ ﺍﻟﺤﻜﻮﻣﻴﺔ ﺍﻟﺘﻲ ﺃﺛﺎﺭﺕ ﺭﺩﻭﺩ ﻓﻌﻞ ﻣﺘﻨﻮﻋﺔ ﻣﻦ ﺍﻟﻤﺠﺘﻤﻊ ﻋﻠﻰ ﻭﺳﺎﺋﻞ ﺍﻟﺘﻮﺍﺻﻞ ﺍﻻﺟﺘﻤﺎﻋﻲ، ﺳﻮﺍﺀ ﻛﺎﻧﺖ ﻣﺸﺎﻋﺮ ﺇﻳﺠﺎﺑﻴﺔ ﺃﻭ ﺳﻠﺒﻴﺔ. ﺗﻬﺪﻑ ﻫﺬﻩ ﺍﻟﺪﺭﺍﺳﺔ ﺇﻟﻰ ﻣﻌﺮﻓﺔ ﺍﺗﺠﺎﻫﺎﺕ ﺍﻟﻤﺸﺎﻋﺮ ﺍﻟﻌﺎﻣﺔ ﺗﺠﺎﻩ ﺑﺮﻧﺎﻣﺞ »ﺍﻟﻮﺟﺒﺎﺕ ﺍﻟﻤﻐﺬﻳﺔ ﺍﻟﻤﺠﺎﻧﻴﺔ« (MBG) ﻭﺗﺤﻠﻴﻞ ﺃﺩﺍﺀ ﺧﻮﺍﺭﺯﻣﻴﺔ IndoBERT ﻓﻲ ﺗﺼﻨﻴﻒ ﺍﻟﻤﺸﺎﻋﺮ ﻓﻲ ﺍﻟﻨﺼﻮﺹ ﺑﺎﻟﻠﻐﺔ ﺍﻹﻧﺪﻭﻧﻴﺴﻴﺔ. ﺃﺟُﺮﻳﺖ ﺍﻟﺪﺭﺍﺳﺔ ﺑﺎﺳﺘﺨﺪﺍﻡ ﺑﻴﺎﻧﺎﺕ ﺍﻟﺘﻌﻠﻴﻘﺎﺕ ﻋﻠﻰ ﻭﺳﺎﺋﻞ ﺍﻟﺘﻮﺍﺻﻞ ﺍﻻﺟﺘﻤﺎﻋﻲ، ﻭﺗﻢ ﻣﻌﺎﻟﺠﺘﻬﺎ ﻣﻦ ﺧﻼﻝ ﻣﺮﺍﺣﻞ ﻣﺎ ﻗﺒﻞ ﺍﻟﻤﻌﺎﻟﺠﺔ ﺍﻟﺘﻲ ﺷﻤﻠﺖ ﺍﻟﺘﻨﻈﻴﻒ، ﻭﺗﻮﺣﻴﺪ ﺍﻟﻜﻠﻤﺎﺕ، ﻭﺍﻟﺘﻘﻄﻴﻊ ﺇﻟﻰ ﺭﻣﻮﺯ، ﻭﺇﺯﺍﻟﺔ ﺍﻟﻜﻠﻤﺎﺕ ﻏﻴﺮ ﺍﻟﻤﻬﻤﺔ، ﻭﺍﺳﺘﺨﻼﺹ ﺍﻟﺠﺬﻭﺭ، ﻭﺍﻟﺘﺼﻨﻴﻒ، ﻭﺃﺧﺬ ﺍﻟﻌﻴﻨﺎﺕ ﺍﻟﻌﺸﻮﺍﺋﻴﺔ ﺍﻟﺰﺍﺋﺪﺓ. ﺃﻇﻬﺮﺕ ﻧﺘﺎﺋﺞ ﺍﻟﺪﺭﺍﺳﺔ ﺃﻥ ﺍﻟﻤﺸﺎﻋﺮ ﺍﻟﻌﺎﻣﺔ ﺗﺠﺎﻩ ﺑﺮﻧﺎﻣﺞ ﺍﻟﻮﺟﺒﺎﺕ ﺍﻟﻤﻐﺬﻳﺔ ﺍﻟﻤﺠﺎﻧﻴﺔ (MBG) ﺗﻤﻴﻞ ﺇﻟﻰ ﺃﻥ ﺗﻜﻮﻥ ﺳﻠﺒﻴﺔ ﻓﻲ ﺍﻟﻐﺎﻟﺐ، ﺣﻴﺚ ﺑﻠﻎ ﻋﺪﺩ ﺍﻟﺒﻴﺎﻧﺎﺕ ﺍﻟﺴﻠﺒﻴﺔ 2948، ﻓﻲ ﺣﻴﻦ ﺑﻠﻎ ﻋﺪﺩ ﺍﻟﺒﻴﺎﻧﺎﺕ ﺍﻹﻳﺠﺎﺑﻴﺔ .1256 ﺑﺎﻹﺿﺎﻓﺔ ﺇﻟﻰ ﺫﻟﻚ، ﺗﻤﻜﻦ ﻧﻤﻮﺫﺝ IndoBERT ﻣﻦ ﺗﻘﺪﻳﻢ ﺃﺩﺍﺀ ﺗﺼﻨﻴﻔﻲ ﻣﻤﺘﺎﺯ، ﺣﻴﺚ ﺗﻢ ﺍﻟﺤﺼﻮﻝ ﻋﻠﻰ ﺃﻓﻀﻞ ﺍﻟﻨﺘﺎﺋﺞ ﻓﻲ ﺍﻟﺴﻴﻨﺎﺭﻳﻮ 10 ﺑﺎﺳﺘﺨﺪﺍﻡ ﻧﺴﺒﺔ ﺍﻟﺒﻴﺎﻧﺎﺕ 80:20، ﻭﺣﺠﻢ ﺍﻟﺪﻓﻌﺔ 16، ﻭﻣﻌﺪﻝ ﺍﻟﺘﻌﻠﻢ 0,00003، ﻣﻤﺎ ﺃﺩﻯ ﺇﻟﻰ ﺩﻗﺔ ﺑﻨﺴﺒﺔ %95,85، ﻭﺩﻗﺔ ﺗﺤﺪﻳﺪ ﺑﻨﺴﺒﺔ %94,42، ﻭﺍﺳﺘﺮﺟﺎﻉ ﺑﻨﺴﺒﺔ %97,46، ﻭﺩﺭﺟﺔ F1 ﺑﻨﺴﺒﺔ .%95,91