Time Series Forecasting Project Presentation.
Download as PPTX, PDF1 like2,267 views
The document outlines a project aimed at forecasting gold prices over the next 30 days using data from 2016 to 2021, with the objective of assisting gold exporting and importing companies in anticipating price movements. It covers various aspects such as data collection, exploratory data analysis, modeling techniques like ARIMA and SARIMA, and the challenges the team faced during development. Ultimately, the project emphasizes the significance of understanding gold price trends and fluctuations for better revenue management.
Time Series Forecasting Project Presentation.
- 2. Contents • Business Objective • Project Architecture • Data Collection & Details • Exploratory Data Analysis • Visualizations • Modeling • Evaluation • Deployment
- 3. Business Objective • Data provided is related to gold prices. The objective is to understand the underlying structure in your dataset and come up with a suitable forecasting model which can effectively forecast gold prices for next 30 days. • This forecast model will be used by gold exporting and gold importing companies to understand the metal price movements and accordingly set their revenue expectations.
- 4. Significance of GOLD GOLD Global Currency Investment Tool Hedging against inflation Stable Commodity
- 5. Project Architecture / Project Flow Business Understanding Data Collection Data Preparation Exploratory Data Analysis Model Evaluation Model Deployment
- 6. Data Collection & Details 2182 rows & 2 column Year range 01-01-2016 to 21-12-2021 Unique Date 2182 and Price 1876 These are the explanations for variables. 1) Date (object) : Daily entry date 2) Price (Float64) : Gold Price
- 7. Importing libraries and Data Sets
- 8. Exploratory Data Analysis (EDA)
- 9. Exploratory Data Analysis (EDA) Information Head Shape Is Null present? Find the Unique values Duplicates Describe the data
- 10. EDA (Visualization) 1.We can see that there is an increasing Trend. So, Trend is not constant. 2.Variance is also not constant. The drastic increase in gold price after the year 2020 and there is intermediate fluctuation in 2021. EDA Visualizations
- 11. Time series decomposition •Trend - Slow moving changes in a time series, Responsible for making series gradually increase or decrease over time. •Seasonality - Seasonal Patterns in the series. The cycles occur repeatedly over a fixed period of time. •Residuals - The behavior of the time series that cannot be explained by the trend and seasonality components. Also called random errors/white noise.
- 12. Visualizing changes in mean over 365 days From the above plot, we can see that there is no constant direction of the mean (increase/decrease) which is understandable as there might be many external factors involved in price fluctuation.
- 13. To Check Normality In The Data
- 15. Bar Plot
- 16. Pie Chart
- 17. Trend & Seasonality in Months
- 18. Trend & Seasonality in Weeks
- 19. Correlation Plot - EDA
- 20. Split The Data
- 23. EDA in HTML using Sweetviz
- 24. Modelling
- 25. Model Building
- 26. ARIMA Model
- 28. Forecast for the 30 Days
- 29. SARIMA Model
- 31. Forecast for next 30 Days
- 32. Holt Method single & Multiple exponential smoothing
- 33. Fit the model tend=‘add', season=‘add'
- 34. EMA Model
- 35. Model Evaluation
- 36. Compering plots ARIMA SARIMA Single exponential smoothing (Hot Encoding) Multiple exponential smoothing ( Hot Encoding)
- 37. Compering values EMA Model Multiple exponential smoothing ( Hot Encoding) Single exponential smoothing (Hot Encoding) ARIMA SARIMA
- 38. Deployment
- 40. Deployment App
- 41. Challenges faced • Research on history of gold prices • Deciding model building technique
- 42. How Challenges Overcome • As we are not much aware of the gold price we have done a lot of research to find out the gold price history. • We as a team worked so hard to get the knowledge on the previous year gold price data. • Which helps us to do the project more effectively.
- 43. Deciding Model Building Technique • As we tried many model building techniques every model don’t have such a significant difference in the output • We are little bit worried about the output results that we got. • But we again overcame this as a team, Everyone has worked really hard on this part and we finally build a model that best suits the data