PUBG吃鸡排名预测案例

因数据集过大不予演示

大致流程

  1. 数据清洗
  2. 模型构建
  3. 模型评估

案例分析：

  1. 获取数据
  2. 数据处理
    去除na，去除开挂人群（无武器击杀，击杀过多，爆头率过高）
  3. 确认x：数据集的特征值
    y：数据集的目标值（标签）
  4. 实例化估计器，进行训练
  5. 进行predict预测，计算score，mae 
  1. import pandas as pd
  2. import numpy as np
  3. from sklearn.metrics import mean_absolute_error
  4. from sklearn.model_selection import train_test_split
  5. from sklearn.preprocessing import StandardScaler
  6. from sklearn.linear_model import Ridge
  8. # 获取数据
  9. from sklearn.preprocessing import StandardScaler
  11. file_path = "./data/train_V2.csv"
  12. df = pd.read_csv(file_path)
  14. print(df.head())
  16. # 基本数据处理
  17. # 去除na
  18. df_new = df.dropna(axis=0, how=any)  # 去除带有na的行
  20. # 选取开挂的数据
  21. df_part = df_new
  23. # 无武器击杀
  24. df_part_distance = []
  25. df_part_distance["weapon"] = (df_part_distance["kills"] > 0) & (df_part_distance["weaponsAcquired"] == 0)
  26. df_part_weapon0 = df_part_distance.drop(
  27.     df_part_distance[df_part_distance["weapon"] == True].index  # 去除无武器击杀的外挂
  28. )
  30. # 击杀过高
  31. df_part_kills = df_part_weapon0.drop(df_part_distance[df_part_distance["kills"] > 30].index)
  33. # 爆头过高
  34. df_part_kills["head_rate"] = df_part_kills["headshot_Kills"]/df_part_kills["kills"]
  35. df_part_kills["head_rate"] = df_part_kills["head_rate"].fillna(0)
  36. df_part_head = df_part_kills.drop(
  37.     df_part_kills[(df_part_kills["head_rate"] == 1) & (df_part_kills["kills"] >= 9)]
  38. )
  40. # 筛选有用的数据作为特征值
  41. df_select = df_part_head[["assists","boosts","DBNOs","heals","kills","matchDuration","rankPoints","revives","rideDistance","teamKills","vehicleDestroys","walkDistance","winPoints"]]
  43. # 确认特征值x和目标值y
  44. x = df_select
  45. y = df_part_head[["winPlacePerc"]]
  46. x_train, x_test, y_train, y_test = train_test_split(df_select.data,df_select.target,random_state=22, test_size=0.2)
  48. # 标准化数据
  49. transfer = StandardScaler()
  50. x_train = transfer.fit_transform(x_train)
  51. x_test = transfer.fit_transform(x_test)
  53. # 实例化估计器 进行训练
  54. estimator = Ridge(alpha=1.0, fit_intercept=True, normalize=False, copy_X=True, max_iter=None,tol=1e-3, solver="auto",random_state=None)
  55. estimator.fit(x_train, y_train)
  57. # 预测值
  58. y_pre = estimator.predict(x_test)
  59. # 准确率
  60. score = estimator.score(x_test, y_test)
  61. # MAE 平均绝对误差（Mean Absolute Error），预测值与真实值的误差绝对值的平均值
  62. mae = mean_absolute_error(y_test, y_pre)