Belarus basketball predictions tomorrow

Market Trends and Public Sentiment Analysis

The betting market often reflects public sentiment towards specific teams or players:

VishalKedia1/crime-prediction<|file_sep|>/README.md # Crime Prediction ## Introduction This project predicts crime rate using Machine Learning models. The data used for this project was collected from [San Francisco Open Data](https://datasf.org/opendata/) website. This data contains information about different crimes that occurred from January year **2003** till June year **2018**. The dataset consists of following fields: - Date : Date when crime occurred. - Category : Type of crime. - Descript : Detailed description of crime incident. - DayOfWeek : Day when crime occurred. - PdDistrict : District where crime occurred. - Resolution : Disposition result. - Address : Street address where crime occurred. - X : Longitude value. - Y : Latitude value. - Location : Location object. ## Dataset The dataset was downloaded from [here](https://data.sfgov.org/Public-Safety/Police-Department-Incident-Reports-Historic-2003/tmnf-yvry). The dataset was downloaded in `.csv` format. ## Data Cleaning To clean up our data we use Pandas library. ### Removing Duplicates The first step we do is removing duplicates from our data. We use `drop_duplicates()` method provided by Pandas library. ### Removing Null Values We remove rows which contain null values from our dataset. ### Removing Invalid Values We remove rows which contain invalid values from our dataset. For example: 1) We remove all rows which contains invalid date values like "01/01/2000". 2) We remove all rows which contains invalid latitude values i.e., latitude values which are not between -90 & +90. 3) We remove all rows which contains invalid longitude values i.e., longitude values which are not between -180 & +180. ### Converting Data Types We convert date column from string type into datetime type using `to_datetime()` method provided by Pandas library. ## Data Preprocessing To preprocess our data we use Scikit-Learn library. ### Feature Engineering We use feature engineering technique called **Binning**. It consists following steps: 1) We group data based on districts i.e., PdDistrict field. 2) Then we divide number of crimes happened in each district into bins based on number of crimes happened there. For example if there were only two bins then we divide number of crimes happened into two parts i.e., less than average number & greater than average number. Then we assign each district one label based on which part it belongs e.g., label '0' if less than average number & label '1' if greater than average number. This will give us one hot encoded vector which will be used as feature later. ### Scaling Features To scale features we use StandardScaler provided by Scikit-Learn library. ### Splitting Data To split data into train & test sets we use `train_test_split()` method provided by Scikit-Learn library. ## Model Training ### K Nearest Neighbors (KNN) KNN is one of simplest Machine Learning algorithms which can be used for both classification & regression tasks. In KNN algorithm it tries to predict value based on its nearest neighbors e.g., if K = N then it returns majority class among N nearest neighbors. ### Decision Tree Classifier Decision tree algorithm tries to find optimal feature & threshold value which gives maximum information gain. It continues until it reaches maximum depth or minimum samples per leaf node. ### Random Forest Classifier Random forest algorithm uses ensemble technique called **bagging** where multiple decision trees are trained independently using different random subsets of data & then averaged out during prediction time. ## Model Evaluation ### Accuracy Score Accuracy score calculates percentage between correct predictions & total predictions made e.g., if model correctly predicts 'n' out of 'm' instances then accuracy score will be n/m*100%. ### Confusion Matrix Confusion matrix provides us information about how many instances were predicted correctly or incorrectly e.g., how many instances were predicted as positive but actually negative etc. ## Results ### K Nearest Neighbors (KNN) Accuracy score obtained using KNN model was **62%** whereas accuracy score obtained using baseline model was **48%**. This shows that KNN model performed better than baseline model. ### Decision Tree Classifier Accuracy score obtained using Decision Tree Classifier model was **72%** whereas accuracy score obtained using baseline model was **48%**. This shows that Decision Tree Classifier model performed better than baseline model. ### Random Forest Classifier Accuracy score obtained using Random Forest Classifier model was **78%** whereas accuracy score obtained using baseline model was **48%**. This shows that Random Forest Classifier model performed better than baseline model. ## Conclusion From above results it is clear that Random Forest Classifier model performs better than other two models i.e., K Nearest Neighbors (KNN) & Decision Tree Classifier models.<|repo_name|>VishalKedia1/crime-prediction<|file_sep|>/code.py import numpy as np import pandas as pd from sklearn.preprocessing import StandardScaler from sklearn.model_selection import train_test_split from sklearn.metrics import accuracy_score from sklearn.metrics import confusion_matrix from sklearn.neighbors import KNeighborsClassifier from sklearn.tree import DecisionTreeClassifier from sklearn.ensemble import RandomForestClassifier df = pd.read_csv('train.csv') # Remove duplicates df.drop_duplicates(inplace=True) # Remove null values df.dropna(inplace=True) # Remove invalid date values df = df[df['Date'].apply(lambda x: len(x.split('/')) == 3)] # Remove invalid latitude values df = df[df['Y'].apply(lambda x: x >= -90 and x <=90)] # Remove invalid longitude values df = df[df['X'].apply(lambda x: x >= -180 and x <=180)] # Convert date column into datetime type df['Date'] = pd.to_datetime(df['Date'], format='%m/%d/%Y %I:%M:%S %p') # Group data based on districts districts = df.groupby('PdDistrict')['Category'].count() # Divide number of crimes happened in each district into bins based on number of crimes happened there. bins = np.array([0,districts.mean(),districts.max()]) # Assign each district one label based on which part it belongs. labels = np.array(['0','1']) # Get one hot encoded vector which will be used as feature later. districts_binned = pd.cut(districts,bins=bins,right=False, labels=labels).astype(int) # Replace district names with labels assigned earlier. df.replace({'PdDistrict':districts_binned.to_dict()},inplace=True) # Scale features scaler = StandardScaler() X = scaler.fit_transform(df[['PdDistrict','X','Y']]) # Split data into train & test sets X_train,X_test,y_train,y_test=train_test_split(X, df['Category'], test_size=0.2, random_state=42) def evaluate_model(model): # Make predictions using trained model y_pred=model.predict(X_test) # Calculate accuracy score acc=accuracy_score(y_pred,y_test) # Print accuracy score print('Accuracy Score:',acc) # Print confusion matrix print('nConfusion Matrix:n',confusion_matrix(y_pred,y_test)) def baseline_model(): # Predict most common class i.e., LARCENY/THEFT. y_pred=[df['Category'].value_counts().idxmax()]*len(y_test) # Calculate accuracy score acc=accuracy_score(y_pred,y_test) # Print accuracy score print('Baseline Model Accuracy Score:',acc) # Print confusion matrix print('nBaseline Model Confusion Matrix:n',confusion_matrix(y_pred,y_test)) baseline_model() model=KNeighborsClassifier(n_neighbors=5) model.fit(X_train,y_train.values.ravel()) evaluate_model(model) model=DecisionTreeClassifier(max_depth=20,max_leaf_nodes=100) model.fit(X_train,y_train.values.ravel()) evaluate_model(model) model=RandomForestClassifier(n_estimators=100,max_depth=20,max_leaf_nodes=100) model.fit(X_train,y_train.values.ravel()) evaluate_model(model)<|repo_name|>VishalKedia1/crime-prediction<|file_sep|>/requirements.txt numpy==1.18.2 pandas==1.0.3 scikit-learn==0.22.2.post1<|file_sep|>#include "Thread.h" #include "Interrupt.h" #include "SystemCall.h" #include "Process.h" extern Process *currentProcess; void Thread::run() { currentProcess->setRunnable(true); while(currentProcess->isRunnable()) { yield(); } } void Thread::yield() { Interrupt::disable(); if(currentProcess->isRunnable()) { SystemCall::schedule(); } Interrupt::enable(); } void Thread::exit() { currentProcess->setRunnable(false); SystemCall::schedule(); }<|repo_name|>csci3308-fall2020/green-goblin-team<|file_sep|>/src/Scheduler.cpp #include "Scheduler.h" #include "Interrupt.h" #include "SystemCall.h" #include "Process.h" extern Process *currentProcess; void Scheduler::schedule() { // Set runnable flag false for current process so scheduler knows not to pick it again before yielding currentProcess->setRunnable(false); // If there aren't any other runnable processes yield back immediately if(runningQueue.size() == 0) { SystemCall::yield(); } // Get next process from queue auto nextProcess = runningQueue.front(); // Set next process runnable flag true so scheduler knows it should pick it again before yielding nextProcess->setRunnable(true); // Remove current process from queue runningQueue.pop_front(); // Add next process back onto queue runningQueue.push_back(nextProcess); // Switch context