<![CDATA[Abstrak Penelitian ini bertujuan untuk mengembangkan sistem rekomendasi tanaman hidroponik berbasis Machine Learning menggunakan metode Feedforward Neural Network (FNN) yang diintegrasikan ke dalam aplikasi HydroSmart. Pemilihan jenis tanaman hidroponik yang tepat memerlukan pertimbangan parameter lingkungan yang kompleks meliputi suhu, kelembapan, pH air, luas lahan, dan intensitas cahaya. Data penelitian bersumber dari platform Kaggle yang divalidasi dengan data resmi Dinas Tanaman Pangan, Hortikultura, dan Peternakan (DTPHP) Provinsi Jambi. Metodologi penelitian mencakup tahapan preprocessing data, perancangan arsitektur jaringan, serta pelatihan model menggunakan pustaka TensorFlow. Hasil eksperimen menunjukkan bahwa model FNN mampu mencapai tingkat akurasi sebesar 91,04% pada data latih dan 90,91% pada data uji. Keunggulan metode ini terletak pada kemampuannya mengekstraksi pola non-linear dari variabel lingkungan secara otomatis tanpa intervensi bobot manual. Model akhir dikonversi ke format TensorFlow Lite (.tflite) untuk memastikan performa optimal pada perangkat mobile. Implementasi ini memberikan solusi cerdas bagi petani hidroponik dalam meningkatkan efisiensi budidaya di berbagai kondisi lingkungan. Kata Kunci: Feedforward Neural Network, Hidroponik, HydroSmart, Machine Learning, TensorFlow Abstract This research aims to develop a hydroponic plant recommendation system using Machine Learning based on the Feedforward Neural Network (FNN) method for the HydroSmart application. Choosing the right type of hydroponic plant requires consideration of complex environmental parameters including temperature, humidity, water pH, land area, and light intensity. Research data was obtained from the Kaggle platform and validated with official data from the Jambi Province Food Crops, Horticulture, and Livestock Service (DTPHP). The research methodology includes data preprocessing stages, network architecture design, and model training using the TensorFlow library. Experimental results show that the FNN model is able to achieve an accuracy level of 91.04% on training data and 90.91% on test data. The advantage of this method lies in its ability to automatically extract non-linear patterns from environmental variables without manual weight intervention. The final model was converted to TensorFlow Lite (.tflite) format to ensure optimal performance on mobile devices. This implementation provides a smart solution for hydroponic farmers to increase cultivation efficiency in various environmental conditions. Keywords: Feedforward Neural Network, Hydroponics, HydroSmart, Machine Learning, TensorFlow]]>
