Data Analysis on some famous datasets available online.
-
Dataset
-
Notebook
-
Implementation Points
- cleaned data to remove uncertainty to ease visualization
- used scattergeo to see the Bombing path, weather station location
- weather data was not staionary as per first Dickey-Fuller test
- used popular methods to get a constant mean
- moving average
- differencing method
- things considered to believe data is now 99% stationary
- by looking at plot , mean looks constant
- variance also looked constant
- test stats from Dickey-Fuller test is smaller that 1% of critical values
- Forecasted Time Series
- used output from differencing method
- used prediction method ARIMA after finding the constant by ACF, PACF
- visualized the ARIMA model prediction and got the mean squared error
-
Dataset
-
Notebook
-
Implementation Points
- Explored Univariate and multivariate timeSeries
- Visualized datasets to understand the Components of TimeSeries
- Trend
- Deterministic Trends
- Stochastic Trends
- Seasonality
- Cyclic Patterns
- Noise
- Trend
- Models for Decomposition of TimeSeries
- Additive Model
- Additive Decomposition
- Multiplicative Model
- Multiplicative Decomposition
- Additive Model
- Visualized the Seasonality in datasets
- TimeSeries Forecasting Techniques
- Moving Average
- Centred Moving Average
- Trailing Moving Average
- Moving Average
- Handling Missing Values
- Forcasting Requirements
- Outliers
- Resampling
- Up-sampling
- Down-sampling
- Measuring Accuracy
- Mean Absolute Error
- Mean Absolute Percentage Error
- Mean Squared Error
- Root Mean Square Error
- ETS ( Error , Trend, Seasonality ) Models
- SES
- Simple smoothing with additive errors
- Holt
- Holt's linear method with additive errors
- Double Exponential Smoothing
- Holt's linear method with additive errors
- Holt-Winters
- Holt Winter's linear method with additive errors
- Multi-step forecast
- Holt Winter's linear method with additive errors
- SES
- Auto Regressive Models
- Auto-Correlation function ( ACF ), Partial Auto-Correlation function ( PACF )
- Stationarity check using Dickey Fuller Test
- ARIMA Model ( AutoRegressive Integrated Moving Average )
- Auto ARIMA
- using AIC ( Akaike Information Criterion ) and BIC (Bayesian Information Criterion ) for model selection
-
Dataset
-
Notebook
-
Implementation Points
- Explored the dataset and visualized missing values
- Used plotly.express to visualize :
- Countried most represented
- Players appearance
- Player's age
- Used plotly.subplots to visualize Players Stats by playing position
- Used plotly.graph_objects to plot graphs for the way goal was made and the mistakes by players
-
Dataset
-
Notebook
-
Implementation Points
- We started with exploring the dataset which gave us an idea about the attributes present in it.
- We later made some changes and visualized the following:
- Carat
- Cut
- Color
- Clarity
- Depth
- Dimensions
- along with their comparision with Price.
- Introduced a new feature Volume to see the relationship between volume of diamond and price.
- Divided the dataset into test and train to use them for evaluating with different Algorithms including :
- Linear Regression
- Lasso Regression
- AdaBoost Regression
- Ridge Regression
- GradientBoosting Regression
- RandomForest Regerssion
- KNeighours Regression
- which gave us the r2 values ( Co-efficient of determination ) and visualized the same which showed us RandomForest Regressor giving the highest r2 value.
-
Dataset
-
Notebook
-
Implementation Points
- Reviwed the train and test data provided
- Calculated the survival rate of men and women
- Predicted suvival rate using RandomForestClassifier