Bayesian Structural Time Series Forecasting with TensorFlow Probability. Structural Time-Series Forecasting with TensorFlow Probability: Iron Ore Mine Production
Iron oreis one of the most heavily traded commodities in the world. As the primary input for the production of steel, it provides the foundation upon which the world’s largest metal market trades, and commands one of the largest shares of the global dry bulk trade.
Iron ore production, unsurprisingly, starts at the mine. As a trader either physical or financial, an understanding of the fundamental supply-demand nature of the iron ore market is essential. Iron ore grade (quality) variance has a notable impact on not only the spot and forward contracts pricing, but also on mill penalty charges for impurities. Imbalances in the fundamental supply/demand relationship can cause dramatic rises in the price of iron ore. Forecasting iron ore output from the largest iron ore exporting countries in order to predict global iron ore supply can be very helpful when speculating on spot, futures and contaminant penalty price movements.
In this article, we are going to develop a forecast model using TensorFlow Probability’s Structural Time-Series (STS) framework, to forecast the aggregate output of major iron ore mines in Brazil.
Brazil is the second largest exporter of iron ore globally. Major changes in supply from Brazil can have an affect on the price of iron ore, as noted above. Furthermore, Brazilian iron ore is typically very high-grade and low in impurities. As a result, the relative supply of ore from Brazil can have an affect on the penalty pricing of impurities charged by steel mills. If global supply is dominated by high-contaminant iron ore as a result of a shortage in supply from Brazilian mines, the price penalty of contaminants can rise dramatically. Forecasting the output from Brazil therefore can lend itself to understanding the above dynamics.
The code used in this article follows similar logic to that outlined in the Structural Time Series modeling in TensorFlow Probability tutorial (Copyright 2019 The TensorFlow Authors).
When approaching a time series forecast problem, investing time into understanding the complexity of the variable you wish to forecast is paramount. Stationarity, seasonality, distributions and exogenous feature relationships are but a handful of the many considerations to bear in mind before designing any model’s architecture.
Structural time series models (sometimes referred to as Bayesian Structural Time Series) are expressed as a sum of components such as trend, seasonal patterns, cycles and residuals:
These individual components are themselves time series defined by a structural assumption. The ability to configure each component in the time series makes TFP’s STS library particularly relevant in the context of our time series forecasting problem, as it enables us to encode domain-specific knowledge, such as trader and mine operator expertise, and known events into our model.
In the programming world, Data types play an important role. Each Variable is stored in different data types and responsible for various functions. Python had two different objects, and They are mutable and immutable objects.
Become a data analysis expert using the R programming language in this [data science](https://360digitmg.com/usa/data-science-using-python-and-r-programming-in-dallas "data science") certification training in Dallas, TX. You will master data...
This Edureka video on 'Python For Data Science - How to use Data Science with Python - Data Science using Python ' will help you understand how we can use python for data science along with various use cases. What is Data Science? Why Python? Python Libraries For Data Science. Roadmap To Data Science With Python. Data Science Jobs and Salary Trends
Python for Data Science, you will be working on an end-to-end case study to understand different stages in the data science life cycle. This will mostly deal with "data manipulation" with pandas and "data visualization" with seaborn. After this, an ML model will be built on the dataset to get predictions. You will learn about the basics of the sci-kit-learn library to implement the machine learning algorithm.
🔥To access the slide deck used in this session for Free, click here: https://bit.ly/GetPDF_DataV_P 🔥 Great Learning brings you this live session on 'Data Vis...