Urban transportation systems are increasingly shaped by high-resolution mobility data generated from mobile devices, smart cards, GPS probes, and shared mobility platforms. While this data abundance enables a shift away from static, survey-based planning toward dynamic, evidence-driven decision-making, extracting policy-relevant insights from high-frequency, spatially heterogeneous mobility data remains a major analytical challenge. This study aims to develop and evaluate a scalable, policy-aligned demand forecasting framework capable of capturing complex urban travel dynamics under real-world conditions. To achieve this, the paper proposes a high-resolution mobility data science framework based on Temporal Convolutional Networks (TCNs). The methodology formalizes the transformation of raw mobility traces into zone-level temporal demand signals, incorporates policy-sensitive features such as peak-period structure and event effects, and applies causal, dilated convolutions to model short- and medium-horizon demand evolution. The framework is evaluated using real-world urban mobility data and benchmarked against econometric (SARIMAX), recurrent neural network (LSTM), and heuristic baselines under a strict out-of-sample evaluation protocol. Results show that the proposed TCN framework consistently achieves lower forecasting error and reduced variance across prediction horizons compared to all baselines, with particularly strong gains under demand volatility and in high-variability urban corridors. The TCN demonstrates greater robustness to non-stationarity, slower error growth with increasing horizon length, and more stable performance across spatial zones. The study concludes that TCN-based high-resolution demand forecasting provides not only superior predictive performance, but also actionable policy insight, enabling earlier identification of demand hotspots, stress-sensitive corridors, and mismatches between supply and usage. These findings position high-resolution mobility data science grounded in temporal convolutional modeling as a practical and scalable decision-support capability for modern urban transportation planning, equity-oriented service allocation, and infrastructure investment prioritization.