{"cells":[{"metadata":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","trusted":true,"collapsed":true},"cell_type":"code","source":"import numpy as np # NUMPY\nimport pandas as p # PANDAS\n\n# DATA VIZUALIZATION LIBRARIES\nfrom matplotlib import pyplot as plt\nimport seaborn as sns\n\n# METRICS TO MEASURE RMSE\nfrom math import sqrt\nfrom sklearn import metrics","execution_count":null,"outputs":[]},{"metadata":{"_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","trusted":true,"collapsed":true},"cell_type":"code","source":"#ALL PUBLIC SOLUTION RMSE < 0.2269 (WITHOUT REPETITIONS)\ndf_base0 = p.read_csv('../input/public-solutions/base0_0.2211.csv',names=[\"item_id\",\"deal_probability0\"], skiprows=[0],header=None)\ndf_base1 = p.read_csv('../input/public-solutions/base1_0.2212.csv',names=[\"item_id\",\"deal_probability1\"], skiprows=[0],header=None)\ndf_base2 = p.read_csv('../input/public-solutions/base2_0.2212.csv',names=[\"item_id\",\"deal_probability2\"], skiprows=[0],header=None)\ndf_base3 = p.read_csv('../input/public-solutions/base3_0.2213.csv',names=[\"item_id\",\"deal_probability3\"], skiprows=[0],header=None)\ndf_base4 = p.read_csv('../input/public-solutions/base4_0.2215.csv',names=[\"item_id\",\"deal_probability4\"], skiprows=[0],header=None)\ndf_base5 = p.read_csv('../input/public-solutions/base5_0.2219.csv',names=[\"item_id\",\"deal_probability5\"], skiprows=[0],header=None)\ndf_base6 = p.read_csv('../input/public-solutions/base6_0.2220.csv',names=[\"item_id\",\"deal_probability6\"], skiprows=[0],header=None)\ndf_base7 = p.read_csv('../input/public-solutions/base7_0.2222.csv',names=[\"item_id\",\"deal_probability7\"], skiprows=[0],header=None)\ndf_base8 = p.read_csv('../input/public-solutions/base8_0.2224.csv',names=[\"item_id\",\"deal_probability8\"], skiprows=[0],header=None)\ndf_base9 = p.read_csv('../input/public-solutions/base9_0.2226.csv',names=[\"item_id\",\"deal_probability9\"], skiprows=[0],header=None)\ndf_base10 = p.read_csv('../input/public-solutions/base10_0.2227.csv',names=[\"item_id\",\"deal_probability10\"], skiprows=[0],header=None)\ndf_base11 = p.read_csv('../input/public-solutions/base11_0.2228.csv',names=[\"item_id\",\"deal_probability11\"], skiprows=[0],header=None)\ndf_base12 = p.read_csv('../input/public-solutions/base12_0.2230.csv',names=[\"item_id\",\"deal_probability12\"], skiprows=[0],header=None)\ndf_base13 = p.read_csv('../input/public-solutions/base13_0.2232.csv',names=[\"item_id\",\"deal_probability13\"], skiprows=[0],header=None)\ndf_base14 = p.read_csv('../input/public-solutions/base14_0.2237.csv',names=[\"item_id\",\"deal_probability14\"], skiprows=[0],header=None)\ndf_base15 = p.read_csv('../input/public-solutions/base15_0.2237.csv',names=[\"item_id\",\"deal_probability15\"], skiprows=[0],header=None)\ndf_base16 = p.read_csv('../input/public-solutions/base16_0.2238.csv',names=[\"item_id\",\"deal_probability16\"], skiprows=[0],header=None)\ndf_base17 = p.read_csv('../input/public-solutions/base17_0.2239.csv',names=[\"item_id\",\"deal_probability17\"], skiprows=[0],header=None)\ndf_base18 = p.read_csv('../input/public-solutions/base18_0.2246.csv',names=[\"item_id\",\"deal_probability18\"], skiprows=[0],header=None)\ndf_base19 = p.read_csv('../input/public-solutions/base19_0.2247.csv',names=[\"item_id\",\"deal_probability19\"], skiprows=[0],header=None)\ndf_base20 = p.read_csv('../input/public-solutions/base20_0.2249.csv',names=[\"item_id\",\"deal_probability20\"], skiprows=[0],header=None)\ndf_base21 = p.read_csv('../input/public-solutions/base21_0.2255.csv',names=[\"item_id\",\"deal_probability21\"], skiprows=[0],header=None)\ndf_base22 = p.read_csv('../input/public-solutions/base22_0.2255.csv',names=[\"item_id\",\"deal_probability22\"], skiprows=[0],header=None)\ndf_base23 = p.read_csv('../input/public-solutions/base23_0.2269.csv',names=[\"item_id\",\"deal_probability23\"], skiprows=[0],header=None)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"collapsed":true,"_uuid":"ebbad7fcd456af13bfb9d18e07886834bd29516e"},"cell_type":"code","source":"#CREATING SOLUTIONS COLUMNS\ndf_base = p.merge(df_base0,df_base1,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base2,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base3,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base4,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base5,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base6,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base7,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base8,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base9,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base10,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base11,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base12,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base13,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base14,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base15,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base16,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base17,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base18,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base19,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base20,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base21,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base22,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base23,how='inner',on='item_id')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"fd3c390de6d180cd2b02f4bf7a7631d6ad8d31fc","collapsed":true},"cell_type":"code","source":"#CORRELATION MATRIX (Pearson Correlation to measure how similar are 2 solutions)\nplt.figure(figsize=(20,20))\nsns.heatmap(df_base.iloc[:,1:].corr(),annot=True,fmt=\".2f\")","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"5053aaace0b1533cb951af99de4f05331c51d5cd","collapsed":true},"cell_type":"code","source":"#ALTERNATIVE WAY - RMSE MATRIX (RMSE to measure how similar are 2 solutions)\nM = np.zeros([df_base.iloc[:,1:].shape[1],df_base.iloc[:,1:].shape[1]])\nfor i in np.arange(M.shape[1]):\n for j in np.arange(M.shape[1]):\n    M[i,j] = sqrt(metrics.mean_squared_error(df_base.iloc[:,i+1], df_base.iloc[:,j+1]))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"9c77a36ac2151f6c5b5eec7f5a71b2fc6be9c6ae","collapsed":true},"cell_type":"code","source":"plt.figure(figsize=(20,20))\nsns.heatmap(M,annot=True,fmt=\".3f\")","execution_count":null,"outputs":[]},{"metadata":{"_uuid":"667b01f7c7f20f4540eb8f2e8f7020baf2fffa39"},"cell_type":"markdown","source":"#Solutions selection:\nWe would like to choose solutions with low correlation and high scores (It is a trade off Correlation vs Score)"},{"metadata":{"_uuid":"5e01d42c3bafc6bfc630171f5031f862e49c7670"},"cell_type":"markdown","source":"**deal_probability0 (0.2211) - Best Solution\n\n**deal_probability1 (0.2212) - Second Solution\n\n**deal_probability2 (0.2212) - Third Solution\n\ndeal_probability3 (0.2213) - Similar to deal_probability0\n\ndeal_probability4 (0.2215) - Similar to deal_probability0\n\ndeal_probability5 (0.2219) - Similar to deal_probability0\n\ndeal_probability6 (0.2220) - Similar to deal_probability0\n\ndeal_probability7 (0.2222) - Similar to deal_probability0\n\ndeal_probability8 (0.2224) - Similar to deal_probability0\n\ndeal_probability9 (0.2226) - Similar to deal_probability0\n\ndeal_probability10 (0.2227) - Similar to deal_probability0\n\ndeal_probability11 (0.2228) - Similar to deal_probability0\n\ndeal_probability12 (0.2230) - Similar to deal_probability0\n\ndeal_probability13 (0.2232) - Similar to deal_probability0\n\n**deal_probability14 (0.2237) - Low Correlation with deal_probability0**\n\ndeal_probability15 (0.2237) - Similar to deal_probability14\n\ndeal_probability16 (0.2238) -Similar to deal_probability0\n\ndeal_probability17 (0.2239) - Similar to deal_probability0\n\n**deal_probability18 (0.2246)- Low Correlation with deal_probability0**\n\ndeal_probability19 (0.2247) - Similar to deal_probability0 and not good Score (0.2247)\n\ndeal_probability20 (0.2249) - Similar to deal_probability0 and not good Score (0.2249)\n\ndeal_probability21 (0.2255) - Low Correlation but Not good Score (0.2255)\n\ndeal_probability22 (0.2255) - Low Correlation Not good Score (0.2255)\n\ndeal_probability23 (0.2269) - Low Correlation Not good Score (0.2269)"},{"metadata":{"trusted":true,"collapsed":true,"_uuid":"ee8527ac424bd440d32a6fc380bf90355b45bec0"},"cell_type":"code","source":"#PORTFOLIO # 0.2204 (0,1,2,14 and 18)\ndf_base = p.merge(df_base0,df_base1,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base2,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base14,how='inner',on='item_id')\ndf_base = p.merge(df_base,df_base18,how='inner',on='item_id')","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"06d4a170b6cc0c403f5668030c5c5a2f4f4b4e27","collapsed":true},"cell_type":"code","source":"#CORRELATION MATRIX (Pearson Correlation to measure how similar are 2 solutions)\nplt.figure(figsize=(10,10))\nsns.heatmap(df_base.iloc[:,1:].corr(),annot=True,fmt=\".2f\")","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"collapsed":true,"_uuid":"fbd3db7d66229888e19b8747816767e4a2a41ac0"},"cell_type":"code","source":"#ALTERNATIVE WAY - RMSE MATRIX (RMSE to measure how similar are 2 solutions)\nM = np.zeros([df_base.iloc[:,1:].shape[1],df_base.iloc[:,1:].shape[1]])\nfor i in np.arange(M.shape[1]):\n for j in np.arange(M.shape[1]):\n    M[i,j] = sqrt(metrics.mean_squared_error(df_base.iloc[:,i+1], df_base.iloc[:,j+1]))","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"_uuid":"30ac4f771ef5606c9e3cc6ab025121eed6997184","collapsed":true},"cell_type":"code","source":"plt.figure(figsize=(10,10))\nsns.heatmap(M,annot=True,fmt=\".3f\")","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"collapsed":true,"_uuid":"a7135f154bf0d0f697c389b86af1fdaf8eb6f249"},"cell_type":"code","source":"#SOLUTION = MEAN OF COLUMNS\ndf_base['deal_probability'] = df_base.iloc[:,1:].mean(axis=1)","execution_count":null,"outputs":[]},{"metadata":{"trusted":true,"collapsed":true,"_uuid":"60b04435eaba9195d0f7bcea91fd153a90eaadab"},"cell_type":"code","source":"#GENERATING FINAL SOLUTION\ndf_base[['item_id','deal_probability']].to_csv(\"best_public_blend.csv\",index=False)","execution_count":null,"outputs":[]}],"metadata":{"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"name":"python","version":"3.6.5","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat":4,"nbformat_minor":1}